From lab to life: technological innovations in transforming cancer metastasis detection and therapy

Abstract

Cancer metastasis remains the leading cause of cancer-related mortality and represents a major therapeutic bottleneck, primarily due to the limited availability of effective, targeted treatment strategies. While key oncogenic drivers such as HER2, EGFR, PIK3CA, KRAS, and BRAF activate critical pathways like PI3K/AKT and MAPK/ERK, promoting tumor proliferation and migration and metastasis. In addition, metastasis is also influenced by environmental factors, microbiomes, and genomic alterations. This complex interplay underscores the urgent need for comprehensive mechanistic insights into metastatic progression, alongside the development of innovative translational platforms. This review explores the external contributors to metastasis, including air and water pollution, chemical exposures, and microbiome dysbiosis, which impact tumor progression and immune evasion. It also discusses the roles of viral infections, organotropism, and genomic regulation in driving metastasis heterogeneity. To address these challenges, a novel integrative framework has been proposed that connects environmental modulators, tumor-associated microbiota, and oncogenic genomic alterations with cutting-edge methodologies such as 3D bioprinting, microphysiological systems, liquid biopsy, and advanced in vitro and in vivo models. High-resolution imaging and AI-driven multi-omics integration further enhance the precision of these approaches. By transcending traditional and reductionist, tumor-centric paradigms, this framework advocates for a systems-level, translational framework that bridges molecular insights with clinical applicability. Ultimately, this strategy seeks to resolve persistent therapeutic challenges in metastatic cancer management through interdisciplinary collaboration.

Introduction

Cancer metastasis is a complex process involving the spread of cancer cells from the primary tumor to distant organs and remains the leading cause of cancer-related mortality. Metastasis involves the dissemination of malignant cells from a primary neoplasm to distant anatomical sites. It is driven by a complex interplay of molecular mechanisms, including chromosomal instability, epigenetic modifications, and dysregulated gene expression, all of which contribute to altered cellular behavior, enhanced invasive potential, and adaptation to new microenvironments. These changes, in combination with tumor heterogeneity and dynamic interactions with the tumor microenvironment (TME), facilitate the ability of cancer cells to detach from the primary tumor, survive in circulation, and establish secondary growths in distant organs. Oncogenes like HER2, EGFR, PIK3CA, KRAS and BRAF enhance cell proliferation, survival, and migration via PI3K/AKT and RAS/RAF/MEK/ERK (MAPK)/ERK pathways [1]. Gain of MYC function promotes angiogenesis and metabolic reprogramming, whereas inactivation of TP53 leads to genomic instability, facilitating metastasis. Epithelial-to-mesenchymal transition (EMT) regulators like TWIST1, SNAIL, and ZEB1/ZEB2 suppress epithelial markers, increasing cancer cell motility [2]. MET and CXCR4 enhance metastasis by promoting cancer cell migration, invasion, and organ-specific colonization. Multiple signaling pathways including PI3K/AKT, WNT/β-catenin, TGF-β, Notch, NF-κB, MAPK/ERK, Hedgehog, Hippo/YAP, and Androgen Receptor (AR) play pivotal roles in tumor survival, EMT, immune evasion, angiogenesis, and therapy resistance across a wide range of cancers, including lung, colorectal, esophageal, pancreatic, liver, ovarian, brain, gastric, breast, and melanoma [3, 4]. Specifically, the PI3K/AKT and NF-κB pathways support tumor cell survival and immune escape, while WNT/β-catenin and TGF-β signaling are key drivers of EMT and metastatic progression. Notch and MAPK/ERK contribute to tumor invasiveness and progression, particularly in breast, colorectal, ovarian, gastric, and melanoma cancers. Hedgehog and Hippo/YAP pathways are closely linked to therapy resistance and metastatic dissemination, especially in pancreatic and liver cancers. In prostate cancer, the AR pathway plays a central role in promoting tumor growth and metastasis [5–7]. Additionally, cancer-associated fibroblasts and immune cells contribute by creating a supportive TME for metastasis. A deeper understanding of these mechanisms is vital for the development of targeted therapies against metastasis [8].

Metastasis diversity refers to the variations observed in how cancer cells spread to different organs, including differences in the number of metastatic sites, the genetic profile of the metastases, the route of spread (bloodstream vs. lymphatic system), and the specific microenvironment of the secondary tumor site, all stemming from the fact that not all cancer cells within a primary tumor have the same capabilities for invasion and colonization in distant organs; this can lead to distinct patterns of metastases even within the same type of cancer [9]. Extensive data across multiple dimensions reveal significant differences in mutational patterns between primary tumors and metastases, as well as among different metastatic sites. This variability influences therapy response and the selective seeding of metastases, as highlighted in recent studies [10]. Such heterogeneity complicates treatment strategies and emphasizes the importance of refining model systems to better capture these nuances.

Addressing metastasis in cancer research remains a formidable challenge, as high clinical trial failure rates stress the inability of current models to fully capture the complexity of metastasis and human diseases, particularly at the molecular level [11]. While advancements in in vitro assays and in vivo animal models have improved our understanding of the molecular and cellular mechanisms driving metastasis, these models still face limitations in robustness, tumor specificity, and clinical relevance [12]. This highlights the urgent need for more predictive and physiologically relevant systems to bridge preclinical findings with clinical outcomes. Continuous innovation is essential to develop more accurate models that better reflect metastatic progression, ultimately enhancing our ability to understand, predict, and treat this critical aspect of cancer.

To resolve this gap, advanced model technologies such as organ-on-a-chip, 3D bioprinting, and AI-driven analysis are essential for improving experimental accuracy and clinical translation. Additionally, precision medicine is crucial in tailoring treatments to individual patient profiles through biomarker discovery, genomic profiling, and therapeutic optimization using tools like CRISPR. Integrating these precision medicine strategies into clinical trials—by focusing on responsive patient subgroups and optimizing trial design—can significantly enhance drug development success rates. In live animals, progressive research in using fluorescent probes, optical imaging, and computational techniques has enabled direct visualization of individual cancer cells during disease progression [13]. Establishment of human cancer cell lines for routine culture, along with the introduction of intravital microscopy (IVM), has allowed for high resolution and detailed observation of metastatic processes within live animal models [14]. Furthermore, precise control over proteomics and protein chemistry, both within cells and through computational modeling, has deepened our understanding of the genetic and molecular mechanisms driving cancer progression [15]. Patient-derived xenografts (PDX) and genetically engineered mice offer better physiological relevance as models but involve species differences and ethical concerns [16]. Computational tools and imaging technologies, like AI-driven models and IVM, enable real-time tracking of metastasis but depend on experimental data, which may introduce biases.

Consequently, efforts are underway to enhance in vitro models to more accurately reflect in vivo conditions. Moreover, the development of drug resistance leads to differences in therapeutic responses between primary tumors and metastases, emphasizing the need for tailored treatment approaches [17]. To develop more effective therapies for advanced-stage cancer, it is essential to create highly refined in vitro models that closely mimic the complexity of the metastatic microenvironment.

Addressing the growing burden of cancer demands a multifaceted approach, encompassing robust cancer surveillance, equitable access to early detection, and continuous advancements in treatment. While reductions in mortality have been seen in high-resource nations, low-resource regions continue to face significant barriers to care, perpetuating global disparities in cancer outcomes [18]. A better understanding of metastasis influencing agents like environmental factors, microbiome, genomic drivers, metastatic diversity and organotropism are essential for developing different experimental tools and detection modalities. With cancer cases and deaths projected to rise significantly in the coming decades, gaining a deeper understanding of tissue-specific metastatic mechanisms will be vital for designing more effective, organ-specific treatments and improving global cancer outcomes in terms of mortality and morbidity.

This review provides a brief overview of different causal relations of metastasis, in vitro functional assays and in vivo models commonly used to investigate metastasis, emphasizing their suitability, limitations, and potential challenges in addressing specific aspects of metastasis research. It also provides technological advancements in studying and detecting metastatic cancer, correlates experimental tools for studying metastasis with other external factors like microbial and viral influence and environmental contributors. Different experimental technologies for exploring cancer metastasis, few clinical trial conditions and role of artificial intelligence (AI) in addressing metastasis have also been discussed. Special focus is placed on recent advancements aimed at enhancing in vivo relevance, lab-on-chip technology for studying metastasis and versatile AI-technologies. This review addresses a critical need to integrate molecular insights, technological advances, and clinical relevance in metastatic research. It distills current knowledge on the drivers of metastasis, surveys a range of diagnostic and modeling platforms, and explores environmental, microbial, and genomic contributors to metastatic spread. It also highlights precision medicine applications, discusses ongoing clinical trials, and provides a future-facing perspective on AI-driven diagnostics and organ-specific modeling platforms. By steering the discussion in specific technologies and pathways, this review aims to support translational research and inform policy directions for more effective, equitable cancer care.

Global status of cancer incidences

Cancer stands second after cardiovascular diseases as a leading cause of mortality in the world. The GLOBOCAN estimated approximately 20 million new cancer cases and 9.7 million cancer-related deaths worldwide in 2022 [19]. This indicates one out of five persons encountered with this deadly disease in life time. Among the cancer types, lung cancer is most prevalent type followed by breast cancer. In the USA, the projected cancer incidence is 2,041,910 with 618,120 cancer deaths in 2025 [20]. The breast cancer in female and prostate cancer in male are most frequent cancer types in the USA. However, the risk factors remain consistent worldwide, including tobacco and alcohol consumption, an unhealthy diet, obesity, and microbial infections such as human papillomavirus (HPV) infection [19, 20]. A study projected a sharp rise in cancer cases and deaths by 2050. Low-Human Development Index (HDI) countries are expected to experience nearly triple the cases and deaths when compared to 2022, while high-HDI countries will see a more moderate increase.

The mortality-to-incidence ratio (MIR) for all cancers in 2022 was 46.6%, with pancreatic cancer showing the highest MIR (89.4%), likely due to its aggressive progression and early metastatic spread to the liver, peritoneum, and lungs. Elevated MIRs were also observed among males (51.7%), individuals aged 75 and older (64.3%), low-HDI countries (69.9%), and the African region (67.2%). The diversity in metastatic behavior, influenced by factors such as tumor genetics, immune response, and organ microenvironments, contributes to these disparities in survival rates [21].

In 2022, lung cancer was the most diagnosed type of cancer, accounting for 12.4% of new cases, and the leading cause of cancer-related deaths, responsible for 18.7% of all cancer deaths. By 2050, it is projected to remain the most prevalent cancer, comprising 13.1% of new cases, with its share of cancer deaths rising to 19.2%, underscoring its ongoing global burden [22]. In 2022, breast cancer (BC) accounted for 2.3 million new cases and 670,000 deaths worldwide. While considering the HDI, though mortality rates from BC have declined in 29 high-HDI countries, incidence rates continue to rise by 1–5% annually in half of the studied nations. By 2050, cases and deaths are projected to increase by 38% and 68%, respectively, with low-HDI countries facing the greatest burden [23]. Oral cancer remains a major concern, particularly in Southeast Asia, where tobacco use significantly contributes to its high prevalence. In India, it accounts for 15% of all cancers, with survival rates remaining poor due to frequent regional and distant metastasis. Cervical lymph node involvement reduces survival by 50%, and metastatic head and neck squamous cell carcinoma (HNSCC) has a particularly poor prognosis. The progression of oral cancer is shaped by metastatic niche formation, where tumor cells modify the microenvironment of secondary sites to support their survival and growth [24]. Processes such as lymphangiogenesis, immune evasion, and extracellular matrix (ECM) degradation—driven by factors like VEGF overexpression, increased expression of matrix metalloproteinases (MMPs) and the loss of E-cadherin—facilitate metastasis [25–27].

Metastasis, is responsible for the majority of cancer-related deaths and continues to pose a formidable challenge in the development of effective therapeutic strategies. It is estimated that metastasis accounts for approximately 90% of all cancer fatalities [28]. A key aspect of metastatic progression is organotropism, which refers to the tendency of certain cancers to metastasize preferentially to specific organs [29]. For instance, colon cancer, which had 1.9 million new cases in 2022, predominantly spreads to the liver, lungs, and peritoneum. This pattern is largely attributed to the rich vascular supply of the liver via the portal vein, facilitating the lodging and growth of circulating tumor cells. In contrast, prostate cancer, with 1.5 million new cases, shows a predilection for metastasizing to bones, followed by the liver and lungs. Notably, this disease demonstrates significantly higher mortality rates among African men, a disparity linked to unequal access to timely diagnosis and advanced treatment options.

Unlike solid tumors, leukemia, a cancer of the blood and bone marrow, disseminates through the bloodstream without forming discrete metastatic lesions. Nevertheless, advances in targeted therapies have markedly improved survival outcomes for patients with hematological malignancies. Meanwhile, brain cancers, such as the highly aggressive glioblastoma, continue to be exceptionally difficult to treat due to the blood-brain barrier and the infiltrative nature of the tumor. Liver cancer, responsible for 760,000 deaths in 2022, is witnessing a rise in global incidence, primarily driven by chronic hepatitis infections and excessive alcohol consumption. Similarly, bone cancer can either originate within the bone or arise as a secondary malignancy from metastases of breast, prostate, or lung cancers, frequently resulting in debilitating pain and fractures. Among skin cancers, melanoma stands out due to its high metastatic potential, spreading rapidly to organs such as the brain, liver, and lungs. Encouragingly, the advent of immunotherapy has significantly improved survival rates for patients with metastatic melanoma. On the other hand, cervical cancer, which often metastasizes to the liver, lungs, peritoneum, bones, and vagina, remains a major global health burden.

Impact of diverse external factors, microbiomes and genomic regulation on metastatic cancer and metastatic organotropism

Environmental factors influencing metastasis

Role of air and water pollution in metastasis

Environmental impact on the incidence of cancer and its metastasis cannot be overlooked, considering the unsustainable industrialization and urbanization. Detrimental effects of air pollution are a growing concern causing respiratory illness and lung diseases [30]. Fine particulate matter-based air pollution is one of the common causes of lung adenocarcinoma. Prolonged exposure to particulate matter (PM) having a diameter of 2.5 μm (PM_2.5) can stimulate tumorigenesis and metastasis of lung adenocarcinoma [31]. Air pollution-induced inflammation drives DNA damage, activates oncogenic pathways, suppresses apoptosis, recruits immune cells, and promotes angiogenesis. Extracellular vesicles (EVs) modulate these processes by transporting bioactive cargo that influences cell signaling, either amplifying, inducing metastasis or mitigating chronic inflammatory responses. In addition, PM of different sizes can be carriers of carcinogenic agents, including organic compounds, heavy metals, which may induce cancer metastasis [32]. Further, EVs are rapidly emerging as early diagnostic biomarkers for tumor progression and metastasis. Thus, detection methods including single EV analysis, e.g., solution-based labelling of EV, multiplexed analysis, digital EV screening and nano-engineered lab-on-chip for the identification of tumor-linked EVs can be employed for promising clinical applications [33, 34].

Emerging evidence highlights a strong association between lung cancer and traffic-related air pollutants (TRAPs), such as particulate matter (PM), nitrogen dioxide (NO₂), ultrafine particles, and elemental carbon. Long-term exposure, particularly to nitrogen oxides (NO_x) and fine particulate matter (PM_2.5), has been linked to both lung and cardiovascular diseases, as demonstrated by several large-scale cohort studies [35, 36]. These airborne pollutants can promote cancer progression by inducing EMT, a key process in invasion and metastasis [37]. In addition to airborne contaminants, environmental pollutants present in untreated industrial effluents discharged into water bodies also pose significant cancer risks. For instance, polycyclic aromatic hydrocarbons (PAHs) adsorbed onto microplastics in seafood have been directly associated with cancer incidence upon human consumption [38]. Furthermore, heavy metals such as arsenic, lead, and cadmium can bind to transforming growth factor-beta (TGF-β) receptors, disrupting critical cellular signaling pathways and contributing to carcinogenesis [39].

Collectively, these findings highlight the multifaceted impact of environmental pollutants, whether airborne or waterborne on cancer development and metastasis.

Role of chemical exposures in metastasis

Exposure to agrochemicals is increasingly recognized as a significant environmental factor contributing to cancer risk, particularly among agricultural workers. Many commonly used pesticides—such as dizion, glyphosate, and parathion—are classified as potential carcinogens and have been associated with tumor progression, immune suppression, and metastatic spread [40]. Occupational exposure to carcinogenic agrochemicals, including organophosphates and organochlorines, is especially hazardous for pesticide-handling farmers, with women showing a strong correlation between such exposure and lymph node metastasis [41]. In breast cancer patients with occupational exposure to pesticides, studies have revealed elevated levels of TGF-β1 and CTLA-4 in both tumor tissues and infiltrating immune cells. These molecular changes reflect impaired immune function and are linked to increased risk of recurrence and mortality [42].

Given the persistent nature of pesticides in soil and water, effective green remediation strategies such as phytoremediation and mycoremediation are essential for sustainable environmental management and reducing long-term ecological and health hazards [43]. Apart from agrochemicals, with the increase in use of plastics globally, microplastics are known to cause conspicuous health hazards due to their persistent and toxic nature. Microplastics are a growing concern for enhancing cancer metastasis in breast cancer [44], wherein polypropylene microplastics significantly increased risks for cancer metastasis by increasing cell cycle progression and IL6 secretion. This demands for detailed, long-term studies on the effect of microplastics in metastasis of other cancers. Additionally, prompt and error-less screening of microplastics is a major challenge faced in present times. Innovative techniques for the detection of microplastics such as atomic force microscopy and atmospheric solid analysis probe can be employed for the effective identification and sustainable mitigation of microplastics in environmental systems [45].

Role of microbiome in metastasis

Mounting clinical evidence underscores the critical influence of host microbiota on cancer susceptibility and metastasis, revealing a complex interplay between microbial communities and tumor progression that is reshaping the current perspectives in oncology. In this regard, Fu et al., 2022 stated that intracellular microbiota was functional in influencing the host-cell actin network and regulating viability in host cells towards tumor progression; tumor-resistant microbiota has a crucial potential in prompting cancer metastasis [46]. Diverse members of the microbiome play a pivotal role in cancer metastasis by shaping the TME, modulating immune responses, and influencing cellular signaling pathways. Microbial imbalance, or dysbiosis, drives key metastatic processes such as EMT, angiogenesis, and immune evasion. In colorectal cancer (CRC), gut dysbiosis fosters a pro-inflammatory environment that compromises the gut vascular barrier. Pathogens like Fusobacterium nucleatum migrate alongside tumor cells to distant sites such as the liver, establishing pre-metastatic niches that facilitate tumor colonization and growth. Studies have identified Fusobacterium sp. in both primary CRC tumors and liver metastases, highlighting its role in metastatic dissemination by influencing the TME, immune responses, and inducing EMT [47].

Intratumoral microbiota further drive metastasis by modulating immune signaling and remodeling the actin cytoskeleton, enhancing cancer cell survival in circulation. For example, Helicobacter pylori has been implicated in EMT induction through cytoskeletal modifications, reinforcing its role in metastatic progression [48]. Studies on microbiota of lung cancer patients suggested that different bacterial genera employ varying processes to cause distant metastasis depending upon the type of cancer [49]. Additionally, changes in the microenvironment during pathological types of cancer influenced the lung microbiome.

These findings collectively emphasize that the microbiota, both within tumors and at distant sites, plays a multifaceted and dynamic role in driving cancer metastasis. Understanding these microbial influences opens new avenues for targeted interventions, where modulation of the microbiome could emerge as a promising strategy to prevent or limit metastatic progression across cancer types.

Viral influences on metastasis

Infectious agents are linked to 10–15% of all human cancers, and viruses play an extensive role in the development of cancer [50]. Oncogenic viruses, like Human Papilloma Virus (HPV), Epstein–Barr Virus (EBV), Hepatitis B and C Viruses (HBV and HCV) and Human T-Cell Leukemia Virus (HTLV-1), are known to trigger carcinogenesis [51]. Common mechanisms employed by DNA oncogenic viruses include targeting tumor suppressors like pRB and p53, integrating viral DNA in host genome and inducing genomic instability, leading to oncogenesis [50]. Some other viruses can induce cell fusion, causing chromosomal instability, which is another common feature of cancer cells [52]. Current research has also found that viruses may play a role in epigenetic modifications by exploiting the epigenetic modifiers of the host for the regulation of viral gene expression, leading to carcinogenesis as a consequence [53].

Viruses exhibit a complex relationship with cancer metastasis, exerting both promoting and inhibitory influences on disease progression. Several oncoviruses—including human papillomavirus (HPV), Epstein-Barr virus (EBV), hepatitis B and C viruses (HBV, HCV), human T-cell leukemia virus type 1 (HTLV-1), and Kaposi’s sarcoma-associated herpesvirus (KSHV)—have been shown to enhance metastasis by promoting anoikis resistance, a key mechanism enabling cancer cells to survive after detachment from the ECM [54]. Since anoikis serves as a natural barrier to metastasis, resistance to it facilitates the survival and dissemination of circulating tumor cells.

In addition, viruses can promote metastasis through various mechanisms such as immune modulation, alteration of cell adhesion molecules (CAMs), induction of inflammation, and stimulation of angiogenesis [55]. Dysregulation of CAMs disrupts normal tissue architecture and enhances tumor cell motility and invasiveness, thereby aiding metastasis [56]. Specifically, the downregulation of E-cadherin, a key epithelial adhesion molecule, by certain viruses plays a pivotal role in initiating EMT, a hallmark of metastatic progression [57]. Moreover, increased expression of pro-angiogenic factors and suppression of anti-angiogenic signals by viral proteins can lead to abnormal blood vessel formation, further supporting tumor growth and the systemic spread of cancer cells [58]. Understanding the multifaceted role of viruses in both tumor initiation and metastasis is crucial for the development of targeted preventive, diagnostic, and therapeutic strategies in virus-associated cancers.

Metastatic diversity and organotropism

Metastatic progression in cancer is governed by distinct molecular mechanisms that vary across tumor types, influencing invasion, organotropism, and therapeutic resistance. For example, breast cancer predominantly metastasizes to the bone, lungs, and liver through EMT regulators such as SNAIL, TWIST1, and the CXCL12-CXCR4 axis [59]. In lung cancer, immune evasion mechanisms involving programmed cell death ligand 1 (PD-L1), regulatory T cells (Tregs), and myeloid-derived suppressor cells (MDSCs) contribute to resistance against EGFR/ALK-targeted therapies [60]. CRC facilitates liver metastasis through ECM remodeling and angiogenesis, mediated by MMPs and TGF-β [61]. Prostate cancer shows a strong predilection for bone metastases, driven by receptor activator of nuclear factor κB ligand (RANKL) and bone morphogenetic protein (BMP) signaling. Pancreatic cancer, on the other hand, enhances survival in hypoxic conditions by leveraging metabolic reprogramming and stemness pathways such as KRAS, WNT/β-catenin, and Hippo/YAP [62, 63]. Melanoma exhibits epigenetic plasticity that enables phenotypic switching and adaptive resistance to B-type rapidly accelerated fibrosarcoma (BRAF)-targeted therapies [64]. Ovarian cancer disseminates within the peritoneal cavity via adhesion molecules and stromal interactions involving Mucin 16 (MUC16) and integrins [65]. Glioblastoma displays highly invasive behavior through MMP-driven disruption of the blood-brain barrier, posing major treatment challenges [66]. Similarly, hepatocellular carcinoma (HCC) promotes metastasis—particularly to the lungs—through VEGF-mediated angiogenesis and immune escape mechanisms involving PD-L1 and WNT/β-catenin signaling [67]. These diverse metastatic mechanisms underscore the complexity of tumor dissemination and highlight the critical need for precision medicine approaches that target both tumor-intrinsic and microenvironmental factors to enhance therapeutic efficacy and improve patient outcomes (Fig. 1).

Fig. 1.

Metastatic diversity and organotropism. This figure depicts certain cancers and the organs to which they commonly metastasise. (a) Breast cancer usually metastasises to organs like the brain, lung, liver and bone, (b) Hepatocellular carcinoma typically metastasises to the lung, (c) Colorectal cancer metastasises to the liver and lung, (d) Prostate cancer commonly metastasises to the bone, (e) Melanoma can metastasise to the brain, lung and the liver, (f) Glioblastoma can metastasise to lung, liver and bone. [Created in https://BioRender.com]

Genomic regulation of metastasis variations

Metastatic progression varies significantly across cancer types and metastatic sites. In pancreatic cancer, the primary tumor develops over approximately 12 years, yet metastases appear rapidly thereafter, with an average onset of 6.8 years [68]. Renal cancer exhibits site-dependent variability in metastasis timing; pancreatic metastases may take a median of 15 years to develop, compared to only 3 years for metastases to other organs. These discrepancies may be attributed to differences in metastatic seeding timing or detection efficiency, as pancreatic metastases often display less aggressive characteristics [69].

Chromosomal fusions, arising from rearrangements or translocations, play a pivotal role in cancer progression and metastasis by modifying gene expression and protein function. These gene fusions can act as oncogenic drivers, sometimes conferring aggressive phenotypes. A study identified six fusion genes across seven human cancers, with CCNH-C5orf30 and TRMT11-GRIK2 found in multiple tumor types—including breast, colon, lung, and ovarian cancers—at frequencies ranging from 12.9 to 85% [70]. These fusion genes were commonly detected in lymph node metastases. While fusion genes such as Breakpoint cluster region–Abelson murine leukemia viral oncogene homolog 1 (BCR–ABL) are known to initiate cancer, metastasis is frequently associated with additional chromosomal aberrations, such as deletions on chromosome 10q25–q26 [71]. The consistent presence of fusion genes in metastatic lesions suggests their continued role in sustaining tumor aggressiveness.

Metastases can emerge early in the disease course, sometimes preceding clinical detection of the primary tumor. In cancers of the pancreas and colon, metastases typically remain genetically similar to the primary tumors, while esophageal cancers tend to disseminate rapidly. The evolutionary patterns of metastasis vary; some maintain close genetic resemblance to the primary tumor, while others undergo significant divergence through parallel evolution. In prostate, renal, and pancreatic cancers, metastases may either diverge early or retain subclones that accumulate additional mutations over time [72, 73].

Different evolutionary mechanisms underpin metastatic dissemination. In CRC, for instance, lymph node metastases demonstrate considerable genetic heterogeneity, whereas distant metastases tend to be more genetically uniform. Some cancers, such as melanoma and prostate cancer, produce distinct subclones capable of seeding multiple metastatic sites. Individual metastases also differ in complexity, originating from single clones or from multiple clones through processes like cross-seeding or aggregation of circulating tumor cell clusters [72].

While metastases typically share the mutational landscape of their primary tumors, they often harbor additional driver alterations. In renal cancer, the loss of chromosome 9p is strongly correlated with metastatic spread [74, 75]. Moreover, whole-genome doubling (WGD) and chromosomal instability (CIN) are common features of metastatic cancers and are associated with poor prognosis. CIN promotes metastasis by amplifying oncogenic signaling, triggering inflammatory pathways, and increasing chromosomal aberrations. Although CIN may result from tumor evolution, evidence suggests it also contributes actively to metastatic progression [76, 77].

Altogether, these insights into the molecular and evolutionary dynamics of metastasis underscore the complexity of tumor dissemination and highlight the necessity for precision therapeutic strategies that address both early detection and the heterogeneous nature of metastatic disease.

Technological advances in detecting metastatic cancer

Conventional modalities

In vitro model systems in cancer research and metastasis research

In vitro models in cancer research provide controlled environments to study behavior of tumor cells, responses to drugs, and communications with surrounding tissues. Therefore, these models serve as essential paradigms to understanding cancer biology and screening potential drug candidates that could be validated in animal studies or clinical trials. In vitro models range from simple two-dimensional cultures to more advanced and complex three-dimensional (3D) structures that closely replicate the TME. Each model has distinct applications, benefits, and limitations, contributing to various aspects of cancer research. The following sections provide a summary of the various in vitro models that could help understand tumorigenesis and tumor invasion [78].

2D monolayer culture The two-dimensional (2D) monolayer culture is commonly used in in vitro studies to understand basic tumor cell proliferation, migration, invasion, and drug screening. Cancer cells are cultured on a flat plastic or glass surface, providing a simple cost-effective, reproducible and manageable making this model a preferred choice for large-scale screening of drugs and mechanistic studies [79]. However, the primary drawback of this model is the absence of a 3D structure, which restricts tumor-stroma interactions and the influence of the ECM. The lack of a heterogeneous tumor environment limits its clinical relevance, as it fails to accurately replicate in vivo tumor behavior [79–81] (Fig. 2).

Fig. 2.

Conventional tools. The figure represents some conventional laboratory techniques and tests used to detect tumor metastasis (a) 2D monolayer culture, (b) 3D spheroid models, (c) Trans-well migration and invasion assay, (d) Hanging-drop assay, (e) In vivo animal models [Created in https://BioRender.com]

3D spheroid models Three-dimensional (3D) spheroid models replicate the tumor environment more effectively than a 2D culture via enhanced cell-cell and ECM interactions. These models have been used quite often to investigate drug resistance, hypoxia, and create a metabolic profile of tumors. Since spheroids are able to generate gradients of oxygen, nutrients and drugs, they provide valuable insights into the physiology of tumors and outcomes of drug treatment [82, 83]. A recent study revealed notable differences in proliferation, gene expression, drug response, and epigenetics between 2D and 3D CRC cultures, with 3D models closely resembling patient samples [84]. However, the lack of vasculature, components of the immune system, and the complex microenvironment, which are observed in actual tumors, continue to limit the utility of this model. Although they show some ECM interactions and heterogeneity, they fail to fully mimic the in vivo tumor complexities [81–83].

Transwell migration and invasion assays The transwell migration and invasion assays are used to measure the motility, migration, and invasion of tumor cells through ECM-coated membranes. These assays facilitate large-scale screening of inhibitors that target the metastatic ability of cancer cells. Their simplicity and cost-effectiveness contribute to their widespread use in cancer research. However, they fail to track invasion in real-time and are unable to mimic the dynamic interactions between tumor cells and the surrounding microenvironment since they lack the 3D complexity. Additionally, they fail to incorporate ECM remodeling and tumor-stroma interactions, restricting their capacity to accurately replicate tumor invasion mechanisms [85]. Figure 2 demonstrates different types of conventional assay platforms.

Hanging drop assay The hanging drop assay has been commonly used to investigate tumor spheroid formation, cell-cell interactions, and the initiation of metastasis. This model facilitates the natural aggregation of tumor cells, forming 3D structures that closely resemble early-stage tumors. Though it is a simple and cost-effective method for studying tumor biology, its limited scalability and absence of tumor-stroma interactions limits its utility for high-throughput drug screening and complex tumor modelling. Despite these drawbacks, the hanging drop assay provides a medium-level complexity model that effectively mimics early-stage tumor aggregates [86] (Fig. 2).

Overall, in vitro models serve as valuable tools in cancer research, each with distinct advantages and limitations. While 2D monolayer cultures enable efficient and rapid drug screening, 3D spheroid models and other advanced techniques provide a more physiologically relevant representation of tumor behavior. A model is chosen based on the specific research question, with ongoing advancements aiming to enhance clinical relevance by integrating key components of the TME, such as vasculature, immune cells, and ECM interactions.

Fluorescence in situ hybridization (FISH) Fluorescence in situ hybridization (FISH) enables the visualization and spatial localization of specific microbial populations within tissue samples, thereby providing critical insights into microbial-host interactions. Advanced variations of this technique, such as catalyzed reporter deposition-FISH (CARD-FISH), enhance sensitivity and allow for the detection of uncultured microbes, while also revealing their morphology and spatial organization in situ [87]. In the context of cancer, FISH analysis of bladder cancer tissues has demonstrated a significantly higher microbial abundance compared to normal tissues, underscoring a potential role of intratumoral microbiota in tumor progression, as reported by Zhang et al. [88].

Polymerase chain reaction (PCR)-based technologies An useful tool for monitoring known metastasis-associated microbes or functional genes is quantitative polymerase chain reaction (qPCR). qPCR assay helps to detect and quantify specific microbial species or genes of interest. It is highly sensitive and specific, and is able to enumerate and detect specific microbial taxa as low as 2.5 × 10³ cells per gram of sample [89]. Similarly, in a study where tumor-resident intracellular bacteria promote metastatic colonization in breast cancer, qPCR was utilized to detect bacterial load in tumor tissue (13 × 10⁵ equivalent bacteria per gram) [45].

In vivo animal models for studying cancer metastasis Naturally existing and genetically engineered models of mice, rats, zebrafish, pigs and dogs have constituted the list of animals employed for studying cancer metastasis, providing valuable insights into the intricate interactions between tumor cells and their microenvironment [90, 91] (Fig. 2). Among the key models, the orthotopic and disseminated metastasis mouse models provide insights into natural tumor growth and rapid assessment of metastasis by replicating the process of invasion from primary tumors and through vasculature, respectively [92–94]. An orthotopic 4T1-Luc breast cancer mouse model enables real-time tracking of tumor growth and metastasis via bioluminescence imaging. After about 30 days, primary tumor removal prolongs survival for metastasis analysis. This model enables tumor progression quantification and anti-metastatic therapy evaluation, making it essential for studying breast cancer metastasis and therapeutics in vivo [93]. In the context of disseminated metastasis, carcinoma cells introduced into lymph nodes of a mouse model rapidly infiltrated blood vessels and formed lung metastases, bypassing the thoracic duct [94]. Likewise, Pereira et al. utilized mouse models implanted with carcinoma or melanoma cells expressing Dendra2 to selectively photoconvert metastatic cells within lymph nodes and monitor their progression. A subset of these cells invaded lymph node vasculature, entered systemic circulation, and colonized the lung [95]. These findings suggest that blood vessels of the lymph node may serve as a route for systemic cancer dissemination, though their significance in human cancer remains to be determined [94, 95].

Syngeneic models involve immunocompetent mice with murine tumor cells to study immune interactions, while xenograft models employ human tumor cells in immunocompromised mice, aiding in the study of biology of cancer [96, 97]. A recent study developed a syngeneic mouse model of HCC driven by Yamaguchi sarcoma viral oncogene homolog (YES), a member of the Src family of tyrosine kinases, which accurately replicated the aggressive metastatic traits for preclinical therapy evaluation [98]. Similarly, recent studies have established syngeneic mouse models of human epidermal growth factor receptor 2 (HER2)-positive metastatic breast cancer, including a novel murine tumor cell line with rapid spontaneous metastasis and an immune-competent model for evaluating anti-tumor immune responses to pharmaco-viral therapy and HER2-targeted treatments [99, 100].

Among models that closely replicate human cancer progression, genetically engineered mouse models (GEMMs) provide valuable insights into metastasis through spontaneous tumor development but are time-consuming and expensive [96]. In this context, the GEMM-embryonic stem cell (ESC) strategy was employed to insert the Met proto-oncogene into a GEMM of BReast CAncer gene 1 (BRCA1)-associated breast cancer, resulting in the creation of a new mouse model of BRCA1-deficient metaplastic breast cancer [101]. Organ-specific models have been used to study metastasis to the brain, bone, lung and liver, offering crucial insights into mechanisms for site-specific metastasis [102]. For example, in a metastatic breast cancer mouse model, HER2 overexpression promoted brain metastasis(BM) progression, tripling large metastases in mice injected with high HER2-expressing cells compared to low HER2-expressing cells [103, 104]. Finally, humanized rodent models incorporate human stromal or immune components, improving relevance for studying human-specific interactions in metastasis, though they require a complex setup and maintenance [105]. SRG-15 humanized mice, incorporating human signal regulatory protein alpha (SIRPA) and interleukin 15 (IL15) knock-ins, facilitate the robust development of natural killer (NK) cells and innate lymphoid cell subsets, enabling efficient antibody-dependent cellular cytotoxicity and NK cell-targeted cancer immunotherapy [106]. In addition, models like intracardiac injection are used to study bone and brain metastases, while intraportal and intraperitoneal injections help to investigate liver and ovarian metastases, respectively [107–109]. A list of the different types of animal models used in the study of metastasis has been provided in Table 1.

Table 1.

The application of different animal models in understanding and studying metastasis

Sl. no	Model	Description	Application	References
1.	Orthotopic metastasis model	Tumor cells are introduced into their original anatomical location where they originated	Investigates metastasis originating from primary tumors	[93, 113]
2.	Disseminated metastasis model	Tumor cells are injected through the tail vein or directly into the heart for wider dissemination. Cancer cells disseminate from the primary tumor to distant sites and form metastatic lesions	Examines metastasis through the vasculature	[92, 94, 95]
3.	Syngeneic metastasis model	Employs immunocompetent mice implanted with murine tumor cells	Enables the investigation of immune interactions in metastasis	[98–100, 114]
4.	Xenograft metastasis model	Utilizes human tumor cells implanted in immunocompromised mice	Investigate tumor growth and metastasis and evaluate anti-cancer therapies	[97]
5.	Genetically engineered mouse model (GEMM)	An immunocompetent mouse model wherein specific genes linked to cancer have been activated or silenced through the use of genetic engineering techniques. Tumors develop naturally in mice	Exhibits spontaneous tumor progression and metastasis in a natural and immune-proficient environment to closely mimic human cancer	[96, 112]
6.	Mouse mammary tumor virus-polyoma middle T antigen (MMTV-PyMT) mammary tumor model (an autochthonous animal model)	Uses a mammary-specific promoter to drive oncogene expression in animals such as mice	Develop spontaneous mammary tumors, rather than being transplanted from an external source (hence, autochthonous) and closely mimics progression and morphology of human breast cancer	[115, 116]
7.	B16 melanoma model	A common melanoma model using B16 cells (melanoma cells) derived from C57BL/6 mice, injected into syngeneic (genetically similar) mice. Examines metastasis of melanoma through tail vein injection	Commonly used to study tumor growth, lung metastasis and develop anti-cancer therapies	[114]
8.	Collagen type I alpha 1-Kremen 2 (Col1a-Krm2) mouse model	A specialized, genetically engineered, and immunocompromised mouse model designed to study how changes in bone microenvironment, specifically driven by WNT signaling inhibition (through KRM2) in osteoblasts, influence cancer metastasis to bone, especially breast and prostate cancer	Investigates breast cancer to bone metastasis	[117]
9.	Intracardiac injection model	The tumor cells are directly injected into the heart of the mouse	Used to study hematogenous (blood-borne) spread of tumor cells. Mimics bone and brain metastases	[109, 118]
10.	Intraportal injection model	Tumor cells are injected into the portal vein under anesthesia, to direct the spread of gastrointestinal tumors to the liver	Used to study cancers like colorectal cancer, and gastric cancer, which commonly spread to the liver via the portal circulation	[119]
11.	Intrasplenic injection model	Tumor cells are injected into the spleen of an anesthetized mouse and is soon followed by splenectomy to allow the tumor cells to enter the liver via portal vein	Used to study liver metastasis, particularly for cancers such as colorectal, and pancreatic cancers	[110, 120]
12.	Intraperitoneal injection model	Tumor cells are injected into the mouse’s peritoneal cavity, where they spread to the peritoneal surfaces, the omentum, and the parathymus	Used to study particularly ovarian cancers, to examine tumor spread in the peritoneal environment and study intraperitoneal chemotherapy	[108, 121]
13.	Humanized mouse model	An immunodeficient mouse engrafted with human immune cells or human tissues	Helps to study human cancer metastasis and immune-evasion mechanisms during metastatic spread in a living organism. Tracks organ-specific human tumor metastasis (e.g., bone, lung, liver). Evaluate human-specific anti-metastatic drugs and immunotherapies [e.g., chimeric antigen T (CAR-T) cells]	[105, 122]
14.	CRISPR/Cas9 gene edited model	It is used to knock out, knock in, or mutate genes involved in tumor invasion, migration, colonization, niche formation, immune evasion, and dormancy	Enables identification of metastasis-regulating genes, modeling of aggressive cancer subtypes, analysis of organ-specific spread and clonal evolution, and in vivo imaging using reporter genes	[123, 124]
15.	Lineage tracing model	A genetic tool used to track the origin and fate of individual cells by employing fluorescent markers during tumor progression and metastasis	Tracking metastatic progression, studying clonal evolution, assessing gene roles in metastasis and identifying early metastatic events before overt spread	[125]
16.	Intracarotid injection model	Tumor cells injected into the common carotid artery of an anesthetized mouse are carried through the arterial blood supply to the brain	Studying brain metastasis in cancers like lung, breast, and melanoma, and understanding how individual gene contributes to brain metastasis and assess the efficacy of therapeutic interventions	[126]
17.	Spinal metastasis model	Tumor cells are injected subcutaneously into the hindlimb of rodents to study spinal metastasis	Studying mechanisms of spinal metastasis, including tumor cell invasion of bone and spinal cord	[127]
18.	Bone disc model	Human bone discs loaded with tumor cells are implanted into the bone cavity of rodents, creating a localized bone metastasis model for tumor-bone interaction studies	Investigating mechanisms of metastasis in bone, including cell adhesion and invasion through the bone matrix	[128]
19.	RNA interference (RNAi ) model	RNAi-based approaches, like short hairpin RNA (shRNA), are used to silence specific genes involved in metastasis. RNAi can be delivered via viral vectors to mice	Helps identify metastasis-related genes	[129]
20.	CRISPR/Cas9 screen model	Cancer cells transduced with pooled CRISPR libraries targeting thousands of genes are introduced into in vivo metastasis models (e.g., orthotopic), and apply antibiotic selection pressure to screen for key regulators of metastasis	Enables researchers to identify key metastasis genes and potential therapeutic targets through targeted disruptions	[130, 131]
21.	Tumor suppressor gene knockout model	A medulloblastoma mouse model can be generated by Cre–LoxP-mediated inactivation of the Rb and Trp53 tumor suppressor genes (TSGs) in cerebellar external granular layer (EGL) cells	Helps to study how genetic ablation of TSGs (Rb in this case) is critical for medulloblastoma development in mice and implicates multipotent EGL progenitors as the tumor cell of origin in the cerebellar EGL	[132]
22.	Human bone marrow model	Primary human osteoblastic cell (hOB)-seeded melt electrospun scaffolds combined with recombinant human bone morphogenetic protein 7 (rhBMP-7) were used to promote a viable humanized bone marrow microenvironment in non-obese diabetic/severe combined immunodeficient (NOD/SCID) mice	This study establishes an in vivo model of human breast cancer metastasis to a humanized bone microenvironment, enabling investigation of the molecular mechanisms underlying tumor homing and growth in bone, as well as therapeutic development	[133]
23.	Spontaneous metastasis model	Tumor cells are orthotopically implanted to form a primary tumor that naturally spreads to distant sites, closely mimicking clinical metastasis	Evaluate the capacity of cancer cells to disseminate from a primary tumor site and are typically initiated via orthotopic injection, wherein cells are introduced into the organ corresponding to the tumor’s tissue of origin	[111, 134]
24.	Allograft model	Tumor cells or tissues from one individual are transplanted into a genetically distinct member of the same species to study cancer growth and metastasis. For example, transplantation of mouse-derived cells into genetically different mice	Investigates metastatic biology and progression. Serves as a useful platform for screening anti-metastatic drugs and development of predictive biomarkers	[135]
25.	Xenograft model with humanized immune system	A human tumor (xenograft) is implanted into an immunodeficient mouse that has been genetically engineered to contain a functional human immune system	Examines tumor-immune interactions for a more accurate representation of cancer progression and treatment response than standard xenografts. They serve as a suitable platform for applications in personalized medicine and cancer immunotherapy	[136, 137]
26.	Metastasis-prone cell line model	A specialized cell line that has the ability to aggressively metastasize and is selected for its high tendency to form secondary tumors in animal models	A robust programmed cell model for investigating epithelial-to-mesenchymal transition (EMT) and invasion in vitro and in vivo, enabling analysis of metastatic mechanisms such as marker expression, signaling pathway activation, chemotherapy resistance, and strategies to inhibit EMT or overcome resistance	[138]
27.	Organ-specific metastasis model	Organ-specific or parental cancer cells are administered via targeted injection routes (e.g., tail vein for lungs, portal vein for liver). The metastasis is monitored via bioluminescence (real-time) and histology/immunohistochemistry (IHC) (endpoints)	Designed to mimic how cancer (e.g., breast cancer) preferentially spreads to certain organs (organotropism), such as lungs, liver, brain or bone. Bone-specific metastatic traits and gene expression patterns observed could potentially be used to understand, predict, or target bone metastasis in human cancer patients	[139]
28.	Tumor-stroma interaction model	A green fluorescent protein (GFP)-expressing immunodeficient nude rat was developed to enable successful engraftment of human and murine tumors, allowing high-purity isolation of stromal cells from xenografts for precise analysis of tumor–stroma interactions	Explores how tumor-stroma interactions drive metastasis by enabling invasion, migration, angiogenesis, and immune evasion. Nude rats with ubiquitous expression of fluorescent markers serve as a unique tool for characterizing the tumor stroma across various human cancers	[140, 141]
29.	Immune cell depletion model	Monoclonal antibodies (mAbs) are used to deplete specific immune cell populations (e.g., CD8+  T cells, macrophages, NK cells) in mice. For example, to deplete CD8+  T cells, WT mice were injected intraperitoneally with anti-CD8β antibody	Examine how missing immune cells, like CD8+  T cells or NK cells, enhance metastasis by enabling cancer to evade immune surveillance	[142, 143]
30.	Metabolic gradient model	Cancer cells are injected into genetically modified mice, and radiotracers could be employed to track their spread toward nutrient-rich areas. For example, proline catabolism fuels lung metastases in breast cancer mice models	Describes how cancer cells metastasize by migrating toward nutrient-rich areas, and targeting these metabolic gradients with therapies or dietary changes could block nutrient use and prevent metastasis	[144, 145]
31.	Ischemia-induced metastasis model	Uses mouse metastasis models that are subjected to ischemic conditions. For example, in a liver metastasis mouse model, hepatic ischemia-reperfusion (I/R) injury is induced as a surrogate for surgical stress during hepatectomies	Investigates the effects of liver I/R on tumor growth and metastasis. This model aids in studying how surgery, bleeding, and septic shock contribute to liver metastasis	[146, 147]
32.	Macrophage-depleted model	Uses genetic or pharmacological methods techniques to deplete macrophages in vivo. For example, KPC mice (expressing activated Kras and inactive Trp53 under the pancreas-specific PDX-1 promoter), which develop tumors that mimic human pancreatic cancer, were administered repeated intraperitoneal injections of clodronate liposomes to deplete macrophages	Investigate how macrophage depletion affects cancer metastasis by examining tumor behavior without macrophage support	[148, 149]
33.	Endothelial cell interaction model	Implants cancer cells in animals to study metastasis and cancer-vasculature interactions in vivo. The interactions of Lewis lung carcinoma cells (both high- and low-metastatic) and murine aorta endothelial cells were studied in mice	Models cancer cell-endothelial interactions in metastasis, focusing on adhesion, signaling, and extravasation	[150, 151]
34.	In vivo barcoding model	Cancer cells are transduced with lentiviral barcode libraries, uniquely labeling individual clones for lineage tracing. These barcoded cells are then implanted into mice (for example, via intracardiac injection), where tumors form and metastasis occurs naturally or are experimentally induced	Enables tracking metastatic tumor origins and uncovering pathways involved in metastasis	[152–154]
35.	Metastasis map (MetMap) model	Creates a map of organ-specific metastatic pattern	Uses barcoding to study metastasis across multiple organs	[154]
36.	Brain metastasis model	Uses intracarotid or intracranial injection to mimic the spread of cancer cells from a primary tumor to the brain in mice	Helps to study the mechanisms and develop treatments for brain metastases, where tumors spread from other body parts to the brain	[155, 156]
37.	Ovarian cancer dissemination model	Ovarian cancer cells (human or murine) are introduced through intraperitoneal injection to mimic peritoneal dissemination. Tumor spread is tracked using bioluminescence imaging	Describes unique spread of ovarian cancer, where cells shed into peritoneal fluid, invading through the mesothelial layer into the organs present in the peritoneal cavity	[157, 158]
38.	CRISPR/Cas9 lineage tracing model	Employs macsGESTALT, an inducible CRISPR-Cas9-based lineage recorder that efficiently captures both transcriptional and phylogenetic information at the single-cell level	Combines CRISPR/Cas9 with lineage tracing. Improves temporal resolution for detecting subclonal molecular changes during metastasis in vivo	[159]
39.	Humanized liver model	A immunodeficient mouse model where human uveal melanoma cells were implanted in its liver through direct hepatic implantation or splenic implantation	To study how human cancer cells behave and spread within a liver environment that closely resembles that of a human patient, particularly in liver metastasis	[160]
40.	Immune-competent mouse model	An adult immunocompetent mice where for example, human glioblastoma (GBM) cells are engrafted and are enabled by transient inhibition of T-cell co-stimulatory signaling	This technique allows GBM tumor growth in immunocompetent mice, closely replicates patient tumor pathology, and serves as a valuable tool for studying tumor biology, immunotherapy, and microenvironmental interactions, with potential applications in other tumor types and models	[161]

Overall, these models offer insights into the role of the TME and genetic contributions to metastasis. They help researchers to monitor tumor progression, evaluate therapeutic interventions, including anti-metastatic therapies, and investigate metastatic mechanisms. This ultimately enhances cancer treatment outcomes, strengthening their indispensable role in cancer research [110]. Despite their usefulness, these models may not completely replicate metastasis of cancers in humans. They require specialized expertise for setup and operation, and some of these models lack interactions with the immune system [109–111]. The integration of in vivo models with emerging technologies, such as CRISPR/Cas9 gene editing and advanced imaging, is expected to significantly enhance our understanding of metastasis [109, 112]. However, rigorous validation and standardization are essential to facilitate their translation into clinical applications, including personalized treatment strategies. As research advances, these models will remain indispensable for driving progress in cancer therapeutics and improving patient outcomes [105, 109, 112].

Recent advancements

Advanced in vitro models for cancer research and metastasis

The development of in vitro models has significantly enhanced our ability to study cancer progression, therapeutic responses, and TME interactions. These models vary in complexity, from simple co-culture systems to highly sophisticated ECM-based 3D scaffolds, biomechanical studies, metastatic niche models, and ex vivo tissue slice cultures. Each system offers unique insights into tumor biology, yet they also present distinct challenges in standardization, reproducibility, and clinical translation. Understanding the strengths and limitations of these models is critical for selecting the most appropriate system for specific research applications [81].

Among the most widely used approaches, co-culture systems simulate interactions between cancer cells and TME components, such as stromal fibroblasts, immune cells, and endothelial cells [162]. These models facilitate the study of tumor-immune evasion, stromal remodeling, and drug resistance mechanisms by replicating signaling crosstalk observed in vivo. For instance, a study developed three optimized and standardized co-culture methods for integrating human macrophages with breast cancer tumoroids (3D patient-derived cancer organoids), using semi-liquid and Matrigel-embedded systems. By refining macrophage numbers and culture conditions, the models effectively captured macrophage-tumor interactions, confirmed through two-photon microscopy and mass spectrometry imaging (MSI). Importantly, macrophage presence influenced tumoroid drug responses, highlighting the need for immune-integrated 3D models to better reflect in vivo tumor behavior. However, standardization remains a challenge due to variability in cell type composition, cytokine gradients, and dynamic cellular interactions, limiting their reproducibility across studies [163].

Notably, tumor spheroid models embedded in hydrogel-based matrices provide a physiologically relevant platform to assess hypoxia-driven adaptations and differential drug penetration [164]. ECM-based models suffer from batch variability in Matrigel composition, limited immune cell integration, and difficulties in replicating organ-specific matrix properties, which can impact experimental outcomes [165].

A critical frontier in metastasis research is the study of the pre-metastatic niche, where distant tissues undergo molecular priming to support future tumor colonization [166]. Metastatic niche mimicking models aim to recreate this environment by exposing naïve organotypic cultures to tumor-secreted exosomes, cytokines, and EVs [167]. However, full recapitulation of the heterogeneous signaling landscape of organ-specific metastatic niches remains a challenge due to the complex interplay of EVs, immune modulation, and matrix remodeling [166, 167].

For translational oncology, ex vivo tissue slice cultures provide an invaluable tool for studying patient-specific tumor biology. These models retain native tissue architecture, stromal interactions, and vascular structures, making them ideal for precision medicine applications and drug sensitivity testing [168]. Recent advancements in organotypic slice culture protocols have enabled multi-day viability, real-time imaging of therapeutic responses, and transcriptomic analysis of drug-induced changes. However, their short lifespan (typically 7–14 days), rapid necrosis at the tissue periphery, and the need for specialized culture conditions limit their widespread adoption [169].

The selection of an in vitro model depends on the complexity of the biological question being addressed. Co-culture systems and ECM-based 3D models provide powerful tools for studying tumor-microenvironment interactions and invasion dynamics, while biomechanical assays offer quantitative insights into metastatic cell mechanics. Metastatic niche models enable investigations into early metastatic seeding and organ-specific tumor tropism, whereas ex vivo tissue slices represent the closest approximation to patient tumors, making them ideal for drug testing. Future advancements should focus on integrating multiple model systems to create hybrid platforms incorporating immune components, ECM heterogeneity, and organ-specific factors. Such innovations will enhance the predictive power of in vitro studies, aligning preclinical discoveries with patient-centered therapeutic strategies in cancer care [166, 167].

Lab-on-a-chip models and microfluidics systems for studying metastasis

Lab-on-a-chip (LoC) technology represents a significant breakthrough in cancer metastasis research, offering unprecedented opportunities for studying complex disease mechanisms, providing physiologically relevant, high-throughput, and scalable platforms that can near-perfectly mimic the TME and metastatic cascade. The rapid progress in micro- and nanofabrication techniques has facilitated the development of sophisticated LoC systems that provide novel platforms for investigating intricate cellular processes [170]. Conventional in vitro and in vivo models often fall short in capturing the intricate spatiotemporal dynamics of metastasis, highlighting the need for microfluidic-based systems that provide precise control over biochemical and biomechanical cues. Innovative platforms such as in vitro microfluidic model for studying microenvironment in oral cancer metastasis, breast cancer-on-a-chip, cancer-on-a-chip, replicate vascularized tumor microenvironments, enabling comprehensive investigations of tumor invasion, intravasation, and therapeutic responses under physiologically relevant conditions. By incorporating key elements such as ECM components, shear forces, and stromal cell interactions, these advanced systems address the limitations of static culture models, paving the way for more predictive and translational preclinical research [170–172]. The integration of LoC technology in breast cancer research has led to significant advancements. Breast cancer-on-a-chip models have been specifically designed to study tumor physiopathology and assess drug efficacy within microenvironments that closely resemble in vivo conditions. Horizontally compartmentalized microfluidic models have proven instrumental in metastasis research by incorporating multiple cell types within a 3D framework to analyze their interactions during cancer progression [173]. The adaptable design of these platforms allows for precise modeling of key metastatic stages, including invasion, intravasation, and extravasation, ensuring controlled experimental conditions for in-depth investigations [170, 172, 174].

Cellular communication plays a pivotal role in metastasis, particularly through EVs such as exosomes and circulating tumor cells (CTCs). Lab-on-a-chip platforms for exosome detection offer real-time, ultrasensitive analysis of exosomal biomarkers, enabling early cancer detection and deeper insights into metastatic progression [175, 176]. To capture the full metastatic cascade, advanced models such as metastasis-on-a-chip (MOC) and cancer metastasis-on-a-chip (CMoC) simulate the sequential steps of metastasis, from tumor detachment and blood-borne dissemination to organ-specific colonization, providing unprecedented insights into metastatic behavior and drug resistance mechanisms [177].

Given the organotropism of metastasis, LoC models have evolved to simulate multi-organ interactions, offering clinically relevant insights into organ-specific metastasis. Multi-organ-on-a-chip systems replicate metastatic progression across multiple tissues, allowing researchers to study the preference of cancer cells for distinct microenvironments. Brain metastasis of non-small cell lung cancer (NSCLC) is a complex process, particularly in breaching the blood–brain barrier (BBB). Due to limited models, its molecular mechanisms remain unclear. Researchers developed a microfluidic chip to replicate BM, identifying AKR1B10 as a key player in BBB extravasation and a potential diagnostic biomarker [178]. Specific models, such as HCC–bone metastasis-on-a-chip and Lung cancer metastasis-on-a-chip, replicate metastatic spread to the liver, bones, and lungs, shedding light on the molecular and mechanical factors influencing organ-specific colonization. A metastasis-on-a-chip to model HCC–bone metastasis included HepG2 cells, a bone-mimetic niche with hydroxyapatite, and a vascular barrier. HCC cells migrated to the bone compartment, with thymoquinone, especially in nanoparticle form, effectively reducing metastasis. This system enables studying cancer progression and drug screening, providing insights into metastasis biology and potential therapies. A microfluidic liver microtissue model using a decellularized liver matrix was established to co-culture kidney cancer (Caki-1) and liver (HepLL) cells. This model demonstrated a strong correlation between Fluorouracil (5-FU) efficacy and cancer cell ratio, with [poly(lactic-co-glycolic acid)-poly(ethylene glycol)] (PLGA-PEG) nanoparticles delivering superior anti-cancer effects [179, 180]. A microfluidic device was developed to study lung cancer metastasis, integrating three chambers: Chamber A (cancer cells), Chamber B (macrophages), and Chamber C (migration monitoring). M2 macrophages in Chamber B upregulated CRYAB (alpha-B crystallin, a small heat shock protein) in lung cancer cells, activating the ERK1/2/Fos-related antigen 1 (FRA-1)/Slug pathway to drive EMT and metastasis in Chamber C. This platform enables real-time monitoring of cancer progression influenced by macrophage-secreted factors [181]. Thus, these platforms facilitate the development of targeted therapies aimed at disrupting organ-specific metastatic niches, thus improving the precision of anti-metastatic interventions [179–181].

Beyond tumor-intrinsic factors, immune and stromal components play a pivotal role in metastasis. Microfluidic models for immune cell interactions, microphysiological systems like Organoid models and Organ-on-a-chip for stromal cell effects allowing researchers to explore how immune cells, fibroblasts, and tumor-associated macrophages modulate metastasis and influence therapeutic response [182, 183]. Furthermore, 3D bioprinted metastasis models and organoid-based metastasis models integrate cutting-edge bioprinting and organoid technology to construct physiologically relevant tumor architectures, enabling detailed studies of EMT, angiogenesis, and tumor dormancy [183, 184]. By serving as an intermediary between conventional 2D cell cultures and in vivo systems, these models significantly improve clinical translatability and support the advancement of personalized therapeutic strategies.

As LoC technology advances, high-throughput metastasis screening models and microfluidic models for angiogenesis are accelerating drug discovery by enabling rapid, cost-effective, and physiologically relevant screening of anti-metastatic agents [185, 186].

Microfluidic devices have also been used to investigate the interstitial fluid flow’s effect on the spread of cancer. For example, it was shown that spheroid-on-chip model that flow conditions might trigger the release of pro-metastatic substances including interleukin 6 and vascular endothelial growth factors [187]. Other studies demonstrate the potential of microfluidic devices to reveal new mechanisms through which the microbiome might influence cancer metastasis by making changes in the TME. A more thorough knowledge of microbiome’s function in cancer metastasis can be found by multi-omics approaches that integrate data from genomics, transcriptomics, proteomics, and metabolomics [188].

Machine learning (ML) enhanced microfluidic detection and classification of metastasis, particularly in biomarker detection underlines the importance of causality analysis in optimizing biomarker detection, TME analysis, material selection, and device design for improved cancer research. The integration of AI and high-content imaging with microfluidic platforms is further refining the precision, automation, and scalability of metastasis research [189]. However, challenges such as technical complexity, standardization, and clinical validation persist. Despite existing challenges, the integration of LoC technologies with AI constitutes a transformative advancement in metastasis research, effectively linking conventional experimental methodologies with the emerging framework of precision oncology. By offering unparalleled control over tumor dynamics, metastatic progression, and therapeutic responses, these technologies hold immense promise for advancing cancer diagnostics, drug development, and patient-specific treatment strategies [172, 189].

Cost-reduction of LoC metastasis models for point-of-care detection requires optimizing design, fabrication, and operational efficiency. Simplifying microfluidic architectures with modular, standardized components can significantly lower production expenses. Using affordable materials like polydimethylsiloxane (PDMS) or thermoplastics instead of costly polymers or glass helps maintain functionality while reducing costs. Additionally, 3D printing enables rapid prototyping and scalable manufacturing without compromising precision [172, 190].

Another strategy involves integrating multi-organ-on-a-chip systems into a single platform to study multiple steps of metastasis (e.g., intravasation, extravasation) simultaneously, reducing the need for separate devices [191]. Automating fluidic systems and incorporating cost-effective imaging techniques, such as fluorescence microscopy, enhance usability while minimizing complexity. Open-source software and AI-driven automation further reduce costs by eliminating reliance on proprietary solutions [192] (Table 2).

Table 2.

Different lab-on-a-chip models for metastasis studies

Lab-on-a-chip model	Working principle	Application	References
Inflammatory-breast cancer-on-a-chip	Collagen type I ECM platforms with an endothelialized blood vessel were cultured with IBC cells (MDA-IBC3 [HER2+] or SUM149 [triple-negative]) and, for comparison, non-IBC MDA-MB-231 (triple-negative) cells. Acellular collagen platforms with endothelialized blood vessels served as experimental controls	3D in vitro vascularized microfluidic platform designed to measure the spatial and temporal dynamics of tumor-vasculature and tumor-ECM interactions specific to inflammatory breast cancer (IBC)	[193]
Lab-on-a-chip for exosome detection	A continuous-flow system for precise isolation and controlled release of blood plasma exosomes across preparation volumes ranging from 10 µL to 10 mL	Microfluidic-based approach (ExoSearch) for efficient enrichment of blood plasma exosomes and in situ multiplexed detection using immunomagnetic beads. Detects ovarian cancer metastasis through exosome analysis	[194]
Cancer metastasis-on-a-chip (CMoC)	3D cultures integrated with microfluidic technology serve as powerful platforms for studying cancer mechanisms and facilitating drug screening	Dissecting the metastatic cascade to enhance understanding of its underlying mechanisms. Current CMoC models replicate key metastatic stages— invasion, intravasation, circulation, extravasation, and colonization—offering valuable applications in drug screening	[177]
Tumor-on-a-chip	Mimics in vivo tumor environments through bioinspired designs and microfluidic technology	Replicates tumor growth, expansion, and angiogenesis while capturing the progression from early to advanced lesions, including EMT, tumor cell invasion, and metastasis	[191]
Metastasis-on-a-chip (MOC)	Cells were perfused through microfluidic channels parallel to lung hydrogel constructs, facilitating direct tumor cell-lung tissue interactions	Thyroid-to-lung metastasis-on-a-chip (MOC) model enabling single-cell-level invasion analysis and quantification	[195]
Hepatocellular carcinoma-bone metastasis-on-a-chip	The bioreactor is comprised of two chambers: one containing encapsulated HepG2 (hepatocellular carcinoma cell lines) and another with a hydroxyapatite (HA)-based bone-mimetic niche. A microporous membrane, simulating the vascular barrier, was positioned above the chambers, enabling continuous medium circulation. Liver cancer cells were observed to proliferate within the tumor microtissue, migrate into the circulatory flow, and eventually infiltrate the bone chamber	To assess the ability of the herb-derived compound thymoquinone to inhibit liver cancer cell migration into the bone compartment	[179]
Lung cancer metastasis-on-a-chip	Includes multiple organ models to study metastasis across different tissues	Investigates the spread of lung cancer to distant organs	[181]
3D bioprinted metastasis model	Combines microfluidics and 3D bioprinting to recreate the tumor microenvironment (TME)	Utilizes 3D bioprinted models to study metastatic progression	[184]
Microfluidic tumor-vascular interface model	A biomimetic tumor-vascular interface by integrating a 3D tumor model with a functional endothelial barrier using a magnetically hybridized system. Endothelial-lined vasculature-like structures on porous microfluidic channels are magnetically connected to tumor spheroids embedded in a collagen matrix within a composite polymer-hydrogel microwell plate	Examines tumor-induced endothelial microdynamics, revealing hallmark changes such as disrupted adherens junctions, increased cell density and proliferation, and endothelial YAP/TAZ nuclear translocation. Established the tumor-vascular interface platform as a valuable tool for the early-stage assessment of anti-angiogenic drug efficacy	[196]
Microfluidic model for epithelial-to-mesenchymal transition (EMT)	Dynamic 3D non-small cell lung carcinoma culture, engineered using multichannel microfluidic platform to assess the impact of flow-induced hydrodynamic shear stress on EMT	The platform enables a dynamic microscale tumor model suitable for early-stage drug screening and treatment monitoring in cancer and related diseases	[197]
Circulating tumor cell (CTC) detection model	A cascaded [Poly(methyl methacrylate) or acrylic] (PMMA) chip-based platform was designed to isolate CTCsfrom blood samples and differentiate between epithelial and mesenchymal CTCs. The primary separation chip (PS-Chip) employed Dean flow fractionation (DFF) to separate CTCs based on size, efficiently eliminating most blood cells while maintaining a high CTC recovery rate of about 90%	The integrated strategy combining efficient enrichment and phenotypic CTC differentiation under high-purity conditions has enabled the cascaded chip’s successful application in diagnosing colon cancer across various stages	[198]
Tumor microenvironment (TME)-on-a-chip	Channel-assembling TME-on-a-chip (CATOC) system, consisting of a blood vessel-on-a-chip and a tumor-on-a-chip, was developed separately by splitting the polydimethylsiloxane (PDMS)-based structure. The blood vessel-on-chip had open windows and a fibrin gel-based basement membrane for endothelial cell culture. Once fully formed, both components were physically connected to assess drug reactivity	Interconnected CATOC enabled evaluation of drug responses across breast TME subtypes, highlighting the value of on-chip assays through comparisons with scaffold-free platforms	[199]
3D Spheroid-based metastasis model	Human (DU 145, MCF-7) and murine (EMT-6) cancer cells were cultured into spheroids for 3, 7, and 11 days. Analysis showed that 11-day DU 145 spheroids exhibited the highest horizontal migration, aligning with RNA-seq data indicating upregulation of cell adhesion, cytoskeleton remodeling, and motility pathways	This model highlights the importance of considering spheroid maturity in cancer research and drug development, highlighting the need for systematic analysis of growth conditions to ensure reproducibility and reliability	[200]
Microfluidic model for angiogenesis	Microfluidic chip system utilized to investigate tumor neovascularization and therapy by co-culturing human-derived endothelial cells with metastatic renal cell carcinoma spheroids. Positioned near primary vessels, the spheroids simulated tumors and triggered neovascular sprouting toward them. Bevacizumab, an antiangiogenic agent, disrupted vessel-tumor interactions, leading to vascular network degradation	A microfluidic chip provided high-resolution analysis of tumor-induced angiogenesis, where endothelial cells sprouted toward the tumor, forming a vascular network. This platform enabled the assessment of endothelial cell biology, vessel functionality, drug delivery, and prostate specific membrane antigen (PSMA) expression	[201]

Liquid biopsy in cancer metastasis detection

Liquid biopsy (LB) has emerged as a promising tool in cancer management, particularly for detecting metastasis. This minimally invasive technique involves analyzing body fluids, such as blood, urine, or cerebrospinal fluid, to identify biomarkers associated with cancer, including CTCs, circulating tumor DNA (ctDNA), and EVs [202, 203]. One of the major applications of LB has been in detecting metastases and monitoring disease progression. They offer several advantages over traditional tissue biopsies, including non-invasiveness, lower risk, and the ability to provide real-time information on tumor status [203, 204].

The concept of LB started gaining attention in the early 2000s when researchers like Kinzler and Vogelstein (1996) provided the first evidence of circulating tumor DNA (ctDNA) in the bloodstream, highlighting its potential as a genetic marker for cancer diagnostics [205]. By the mid-2000s, advancements in technologies such as PCR and next-generation sequencing (NGS) significantly improved the sensitivity of ctDNA detection in circulation [206]. A 2005 study evaluated serum DNA methylation for early lung cancer detection by analyzing the methylation status of five tumor suppressor genes in 200 patients undergoing bronchoscopy. Methylation was detected in 50.9% of stage I lung cancer cases, compared to only 11.3% positivity with serum protein tumor markers, suggesting the potential of serum DNA methylation as an early detection tool [207]. Subsequently, a study by Maheshwaran et al. demonstrated that CTC scan be isolated from lung cancer patients’ blood to non-invasively monitor EGFR mutations, including drug resistance mutations like T790M, with CTC count changes correlating with treatment response and tumor progression [208]. Again, Guo et al. found that ctDNA levels dropped in 91.7% of cases post-surgery, with detectable changes within two days, and showed a higher predictive value than six clinical tumor biomarkers for tracking treatment response and disease progression in NSCLC patients [209]. Another study published in 2020 optimized a LB workflow using cell-free DNA (cfDNA) and CTCs from a single blood sample to monitor EGFR mutations in NSCLC patients. Results demonstrated high sensitivity, enabling non-invasive therapy selection and treatment monitoring, with promising accuracy compared to traditional tumor biopsies [210]. A multi-institutional study of 1487 oligometastatic NSCLC patients found that pre-radiation therapy (pre-RT) ctDNA levels predict progression-free and overall survival. Undetectable ctDNA correlated with improved outcomes, aiding in risk stratification for locally consolidative radiation therapy versus systemic therapy for micrometastatic disease [211]. Thus, in lung cancer, liquid biopsies can detect metastases with high accuracy, even before they are visible on imaging scans, allowing for early intervention and personalized treatment [212].

Essential Biomarkers in LB:

1. Circulating tumor DNA (ctDNA): ctDNA, released by tumor cells into the bloodstream, carries genetic mutations and epigenetic alterations essential for early cancer detection, monitoring treatment response, and identifying minimal residual disease (MRD). Its advantages include enabling real-time tumor genotyping, tracking drug resistance, and predicting relapse. However, its limitations include low sensitivity in early-stage cancers and challenges in differentiating ctDNA from normal cfDNA [213]. The United States Food and Drug Administration (FDA) has approved Guardant Health’s Guardant360 CDx and FoundationOne Liquid CDx as commercial platforms that utilizes ctDNA analysis for diagnosis and monitoring of breast cancer, NSCLC, and advanced solid tumors [212, 214, 215]. The CellSearch system (Menarini Silicon Biosystems) is an FDA-approved CTC detection method for patients with metastatic breast, prostate, or CRC. It selectively enriches cells that express epithelial cell adhesion molecules (EpCAMs) while lacking the leukocyte common antigen CD45 [216]. Similarly, Natera’s Signatera assay monitors minimal residual disease (MRD) by tracking patient-specific tumor mutations over time, offering insights into relapse risk and treatment response [217].

2. Cell-free DNA (cfDNA): cfDNA, originating from both tumor and normal cells, enables the analysis of genetic alterations, aiding in early cancer detection. Although cfDNA offers a minimally invasive approach for genetic evaluation and can enhance ctDNA analysis, the high background levels of non-tumor cfDNA may dilute signals derived from the tumor [212].

3. DNA fragmentomes: The fragmentation patterns of cfDNA, known as DNA fragmentomes, provide valuable insights into tumor biology and presence. While not yet widely adopted in clinical practice, DNA fragmentome analysis can supplement mutation analysis and aid in identifying tumor subtypes. However, its interpretation requires advanced computational tools [218].

4. MicroRNA (miRNA): These small RNAs regulate gene expression and contribute to tumorigenesis and metastasis. Their stability in circulation makes them reliable biomarkers for early cancer detection and tumor subtype identification. However, differentiating tumor-derived miRNAs from other sources remains challenging [219].

5. Exosomes: Tumor-derived EVs are exosomes, which carry molecular information that reflects the TME. Tumor-derived exosomes (TDEs) play a crucial role in tumor metastasis by facilitating communication between tumor cells and normal cells, promoting migration, angiogenesis, immune suppression, and tumor implantation. They show potential for early cancer detection and monitoring treatment response. However, challenges remain in standardizing their isolation and detection due to their small size and low abundance in circulation [212].

6. Tumor-educated platelets (TEPs): Cancer-altered platelets can act as biomarkers for detecting tumor presence. Tumor-educated platelets (TEPs) enable highly sensitive cancer detection and subtype differentiation, though the mechanisms behind platelet-tumor interactions are still being studied [220].

7. Protein biomarkers: Proteins derived from tumors, such as carcinoembryonic antigen (CEA) and prostate-specific antigen (PSA), are widely used in clinical cancer diagnostics. However, their limited specificity and sensitivity, especially in early-stage cancer, reduce their effectiveness for comprehensive detection [221].

Although LB holds great potential as a minimally invasive means to detect metastasis and monitor disease progression, challenges remain, including detection sensitivity and the need for standardized protocols [204]. Future research should prioritize integrating LB with advanced technologies like AI and machine learning to improve predictive modeling and personalized medicine [203].

Several companies are at the forefront of liquid biopsy and molecular testing, advancing early cancer detection, diagnosis, and monitoring. For example, a company named GRAIL is conducting the SUMMIT study to evaluate a blood test for detecting multiple cancers, including lung cancer. Another company, Guardant offers Guardant 360 and Lunar-2 for early cancer detection in high-risk, asymptomatic individuals, while Freenome leverages AI-driven screening to identify CRC and precancerous lesions. Biocept’s Target Selector™ ctDNA EGFR Kit specializes in EGFR mutation detection, and Inivata’s InVisionSeq™ and InVisionFirst™-Lung provide a 40-biomarker panel for molecular profiling, monitoring, and diagnosing advanced cancers. Cynvenio’s LiquidBiopsy^® Platform employs NGS-based techniques to detect mutations from as few as one target cell per mL of blood. CellMaxLife’s FirstSightCRC™ utilizes CTCs for CRC screening, while ExosomeDx focuses on exosome-based biomarker tests for diagnosing non-small cell lung and prostate cancer. Biodesix’s GeneStrat^® test delivers blood-based mutation results for EGFR-, ALK-, ROS1-, RET-, BRAF-, and KRAS-encoding genes, and Personal Genome Diagnostics’ PlasmaSELECT™-R 64 uses an NGS-based 64-gene panel for cancer diagnosis. These cutting-edge innovations are transforming non-invasive cancer detection, enabling earlier intervention and more targeted treatment strategies [222].

Single-cell RNA sequencing (scRNA-seq)

Single-cell RNA sequencing (scRNA-seq) is a powerful tool for investigating the TME and cancer metastasis at high resolution. It also enables the study of microbial population heterogeneity and host–microbiota interactions at the level of individual cells. Robinson et al. (2024) employed a computational pipeline known as Computational Identification of Cell type Specific Intracellular Microbes (CSI-Microbes) to identify microbial reads from scRNA-seq data and to analyze the differential abundance of taxa and intracellular polymicrobial interactions. Their findings revealed that myeloid cells containing engulfed bacteria were the primary sources of bacterial RNA within the TME, suggesting a significant role in modulating immunotherapy response and inflammation. In colorectal and esophageal cancers, tumor cells infected with bacteria were observed to upregulate antigen processing and presentation pathways. These unanticipated cancer cell–microbiome interactions imply that microbes may influence cancer cell behavior and potentially impact metastasis [223].

Metatranscriptomics

Metatranscriptomics, a technique involving the sequencing and analysis of RNA molecules from microbial communities, enables researchers to investigate active metabolic pathways and gene expression patterns. This approach has uncovered associations between specific bacterial genera and metastatic stages of CRC, revealing distinct signatures related to immune regulation and cancer metabolism through metabolomic profiling [224]. In prostate cancer, gene expression from Pseudomonas species has been correlated with host gene expression, suggesting a potential influence on metastasis [225]. Furthermore, metatranscriptomic studies have demonstrated that bacteria can play dual roles during inflammation—some promoting pathogenesis, while others help maintain physiological homeostasis [226]. In pancreatic cancer, metatranscriptomics has shown promise for identifying novel biomarkers that could improve early diagnosis and prognosis of metastatic disease [227]. These findings highlight the complex interplay between the microbiome and cancer development, emphasizing the need for further research to fully elucidate these mechanisms.

Understanding the mechanisms of oncovirus-dependent metastatic spread utilizing CRISPR-Cas9 technology

A better understanding of virus–host interactions and the mechanisms by which oncoviruses contribute to cancer and metastasis is essential for developing new treatments and therapies to combat metastatic disease. The advent of CRISPR-Cas9 technology has revolutionized the study of these interactions, particularly in cancer research. This tool enables genome-wide investigations to systematically identify host-dependent factors critical for viral pathogenicity and replication by mapping complex interactions and allowing targeted disruption of specific host genes involved in viral infection [228]. CRISPR screening has been applied to clinically relevant viruses such as Zika virus, West Nile virus, hepatitis C virus (HCV), and dengue virus, yielding valuable insights into virus–host dynamics [229]. Additionally, CRISPR-Cas9 has demonstrated promising potential in targeting host factors and viral genomes to eliminate viral reservoirs and prevent infection and replication, offering therapeutic promise in gene therapy research [230]. Beyond virology, it is also used in establishing tumor models, screening potential immuno-oncology targets, and manipulating the cancer genome and epigenome. However, despite these advantages, limitations such as editing efficiency, fitness of edited cells, off-target effects, and delivery challenges must still be addressed to fully harness the potential of CRISPR-Cas9 technology [231].

Genome sequencing technology in detecting metastasis and therapy

Detecting and understanding the genomic alterations that drive metastasis is crucial for advancing personalized medicine and improving patient outcomes [232]. Genome sequencing technologies have emerged as powerful tools to unravel the molecular landscape of metastatic tumors, enabling precise diagnosis, prognosis, and therapeutic interventions [216, 233, 234].

Recent advancements in sequencing methodologies have facilitated the comprehensive analysis of metastatic cancer genomes, identifying critical mutations, copy number variations, and gene expression alterations that contribute to tumor spread [235]. Robinson et al. analyzed 500 metastatic tumors, identifying key alterations in genes such TP53, CDKN2A, PTEN, PIK3CA, and RB1 and highlighted the value of integrative sequencing in understanding the complex molecular landscape in metastases. Techniques such as Whole-Genome Sequencing (WGS) (Platforms: Illumina HiSeq^®, NovaSeq^®) and Whole-Exome Sequencing (WES) (Platforms: Illumina TruSeq^®, Agilent SureSelect^®) provide deep insights into the genetic basis of metastasis, while targeted sequencing focuses on specific metastatic drivers for efficient clinical decision-making [236–241].

Whole-exome sequencing of 154 tumor-normal pairs from 97 patients with metastatic cancers revealed an average of 16 somatic alterations per patient, with 114 mutations classified as biologically relevant or potentially actionable; however, treatment was influenced in only 5% of cases due to limitations in clinical trial access and off-label drug use. A notable discovery was a prostate cancer patient with a deletion in FANCA (a gene involved in Fanconi anemia), who showed exceptional sensitivity to platinum-based chemotherapy, suggesting the potential of WES in identifying predictive biomarkers for treatment response [236]. An analysis of 7,108 whole genome-sequenced tumors across multiple cancer types revealed that metastatic cancers exhibit lower heterogeneity, stable karyotypes, and increased structural variants, with certain cancers undergoing significant genomic changes and treatment driving resistance selection [238]. Next-Generation Sequencing (NGS) (Platforms: Illumina MiSeq^®, Ion Torrent^®) further enhances these capabilities of studying metastatic cancer by enabling high-throughput analysis with remarkable sensitivity and speed [242–244]. Targeted next-generation sequencing (tNGS) platforms, such as Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT^®) and FoundationONE^® verified that RAS mutations were present in both primary and metastatic CRC samples. Thus, these platforms enabled the reliable identification of clinically actionable mutations, improved outcomes in subsets of cancer patients through matched therapies, revealed stable and de novo mutations across disease stages, and demonstrated the power of comprehensive genomic profiling to guide precision oncology and make informed therapeutic decisions [245–247]. Furthermore, unique de novo mutations were also detected in synchronous and metachronous lung and liver metastases, with some late-stage mutations in metachronous lung metastases potentially treatable with targeted therapies [246]. In addition to DNA-based methods, copy number variation (CNV) analysis (Platforms: GISTIC2.0, OncoScan^®) and genome-wide RNA interference (RNAi) screens [Platform: Dharmacon siRNA Libraries™ from Dharmacon (Horizon Discovery)] provide insights into genomic instability and functional gene interactions driving metastatic processes [246, 248–252]. For instance, the computational pipeline for CNV detection in NGS data successfully identified CNVs in 36 positive control samples with 100% sensitivity and 91% specificity, enabling the detection of whole-gene, exonic, and partial exonic variations. Since its introduction in 2018, it has enhanced diagnostic capabilities by enabling CNV detection across all target panel genes, addressing the prior limitations of multiplex ligation-dependent probe amplification (MLPA) kits [253].

Additionally, microarray and gene chip technologies (Platforms: Affymetrix GeneChip^®, Agilent Microarray^®) provide a broader view of gene expression changes associated with tumor dissemination and recurrence [254–256]. A metastasis-related gene signature was analyzed in two independent HCC cohorts (totaling 386 patients) and demonstrated its ability to predict overall and disease-free survival, regardless of clinical features and microarray platforms. It was especially effective in detecting early-stage patients at high risk of recurrence within two years, particularly when combined with serum alpha-fetoprotein or tumor staging, reinforcing its potential as a molecular diagnostic tool for assessing relapse risk [255]. Microarray-based gene expression profiling has significantly improved metastasis detection across various cancer types, exemplified by the MammaPrint assay—a 70-gene signature that evaluates breast cancer recurrence risk, enabling clinicians to personalize treatment decisions [257]. In melanoma, the CP-GEP model combines clinicopathologic (CP) factors with gene expression profiling (GEP) to evaluate nodal metastasis risk, potentially reducing the need for sentinel lymph node biopsies [258]. Likewise, microarray analyses in CRC have identified differentially expressed genes linked to metastasis, providing valuable insights into tumor behavior and therapeutic targets [259]. Large-scale meta-analyses across cancers have uncovered common metastatic pathways and biomarkers, deepening our understanding of metastasis [260]. Collectively, these technologies play a crucial role in risk assessment, prognostication, and elucidating molecular mechanisms of metastases, ultimately guiding treatment strategies and enhancing patient outcomes [261]. With the ongoing integration of sequencing technologies with AI and machine learning, the future of metastatic cancer diagnostics and treatment appears highly promising [262]. Utilizing these advanced genomic platforms, clinicians and researchers can enhance precision oncology approaches, leading to improved patient survival and quality of life [263].

Detection of microbial fingerprints associated with metabolic pathways associated with cancer development and treatment outcomes can be explored using NGS. Combining NGS with AI and machine learning allows for a personalized treatment approach, by enabling the development of diagnostic tools to cater to a patient’s microbiome profile [264].

16S rRNA gene sequencing and whole genome shotgun metagenomic sequencing (WGSMS) have contributed significantly to our understanding of gut microbiome and its potential influence on cancer metastasis [265]. 16S rRNA gene sequencing focuses on a specific region of the bacterial genome, while WGSMS provides a comprehensive view of the entire microbial community [266]. Both methods have been used to explore the relation between CRC and gut dysbiosis and identify microbial signatures associated with CRC development. Parvimonas micra is one such organism, whose link to CRC has been identified using these techniques [267]. Although WGSMS can often provide more details, studies have shown that 16S rRNA data can yield similar results in the cases of alpha diversity, beta diversity and disease prediction accuracy, especially in cases of paediatric ulcerative colitis [268].

In summary, the integration of microarray, gene expression profiling, and NGS technologies—augmented by AI and machine learning—has transformed metastatic cancer diagnostics. These approaches enable early risk prediction, personalized treatment planning, and the identification of key molecular and microbial drivers of metastasis. Tools like 16S rRNA and WGSMS further enhance our understanding of the microbiome’s role in cancer progression, paving the way for more precise and effective oncological interventions.

Metabolomic tools for cancer metastasis

Cancer metabolism is highly diverse, varying across tumor regions due to genetic mutations and microenvironmental factors. Several models and technologies are employed to investigate these variations. In vitro 2D cultures are widely used to assess metabolic heterogeneity using tools like the Seahorse Flux Analyzer and metabolic flux analysis, which measure real-time changes in cellular respiration and glycolysis [269]. Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) provides spatial metabolic profiling of tumor tissue sections, allowing for the visualization of metabolic differences within the tumor [270].

Single-cell metabolomics (SCM) is revolutionizing cancer research by enabling precise characterization of metabolic heterogeneity within tumors. Unlike bulk metabolomics, which provides an averaged metabolic profile, SCM identifies distinct metabolic states of individual cancer, immune, and stromal cells, offering crucial insights into tumor heterogeneity, metastatic adaptations, and therapy resistance. A single-lysosome mass spectrometry (SLMS) platform was developed, integrating lysosomal patch-clamp recording with nano-electrospray ionization mass spectrometry (nano-ESI-MS) for simultaneous metabolic and electrophysiological analysis. Validated by prior findings, SLMS identified five lysosomal subpopulations and revealed metabolic alterations in senescence and carcinoma. This technique advances research on lysosomal metabolism in health and disease [271]. SCM has been instrumental in metastatic melanoma and head and neck cancers, uncovering metabolic differences between tumor and non-tumor cells that remain undetectable in bulk analyses. Its potential extends to CTCs, serving as a liquid biopsy biomarker for metastasis. Despite several challenges, advancements in sampling protocols, ionization techniques, and metabolite detection are continually refining SCM’s accuracy. Further optimization of cell isolation and metabolic preservation will be crucial in enhancing its application in cancer diagnostics, treatment response assessment, and metastasis research [272].

Quantitative metabolomics has long been a vital tool for analyzing metabolites in biological samples. Like other ‘-omics’ sciences, metabolomic technologies continue to evolve, driving advancements in analytical techniques, models, software, and computational methods to enhance sensitivity and specificity. Established techniques such as nuclear magnetic resonance (NMR) spectroscopy, gas chromatography–mass spectrometry (GC-MS), and liquid chromatography–mass spectrometry (LC-MS) have been instrumental in detecting and quantifying metabolites associated with cancer metastasis. However, these methods typically require a large number of cancer cells (> 10,000) and lack the resolution to assess metabolic heterogeneity at the single-cell level. Understanding metabolic variations within a tumor is crucial, as distinct subsets of cancer cells may exhibit differing metastatic potential and treatment responses. Identifying specific metabolic traits in these subsets could provide valuable biomarkers for predicting metastatic efficiency and informing diagnostic and prognostic strategies.

Metabolomic profiling using mass spectrometry (MS) and NMR spectroscopy are being used to detect microbial metabolites that could influence cancer progression and metastasis by identifying biomarkers associated with microbial activity. In cancers like CRC, microbiota sequencing and advanced metabolomics studies have helped to identify characteristic patterns in gut microbiome and faecal metabolites across different metastatic sites [273]. The role of the microbiome extends beyond the gut. In breast cancer, metabolomics has shown links between microbiome expression and metabolic dysregulation. Breast tumor cells were characterized by elevated glycolysis and tricarboxylic acid cycle and were highly sensitive to microbiota activity [274]. Studies conducted using NMR-based metabolomics have identified unique metabolic profiles distinguishing cancer patients from healthy individuals. In the case of CRC, ¹H-NMR analysis of urine samples revealed as many as 16 metabolites that could function as potential biomarkers in early-stage CRC. These indicated disruption of metabolic pathways such as glycolysis, amino acid metabolism and gut microflora metabolism [275].

Evolution of cancer detection imaging: from classical methods to cutting-edge technologies

Over the years, the approaches to detect metastatic cancer has evolved from the classical imaging such as computerized tomographic (CT) scans and magnetic resonance imaging (MRI) followed by positron emission tomography (PET) to the extremely advanced molecular diagnostic, machine learning (ML) and computational methods [276]. Each advancement has addressed specific limitations of its predecessors, leading to more accurate and less invasive diagnostic methods [276, 277]. Introduced in the early 20th century, X-rays provided the first non-invasive means to detect bone abnormalities and certain tumors; however, their limited soft tissue contrast necessitated the development of more sophisticated imaging techniques [278]. The advent of nuclear medicine in the 1970s introduced the use of radioactive tracers to assess physiological functions, with bone scans becoming instrumental in detecting metastases [279]. Ultrasonography emerged as a radiation-free method to visualize internal structures, proving especially useful for detecting superficial tumors in organs such as the thyroid and breast, and its real-time imaging capability made it invaluable for guiding biopsies and other interventional procedures [278, 280]. The development of computed tomography (CT) combined X-rays with computer processing to offer cross-sectional images of the body, enhancing tumor detection and characterization [277]. Magnetic resonance imaging (MRI) utilized magnetic fields and radio waves to produce detailed images, particularly of soft tissues, making it essential for evaluating brain tumors and other soft tissue malignancies [281]. Positron emission tomography (PET) assessed metabolic activity by detecting radioactive tracers, aiding in the evaluation of tumor metabolism and the detection of metastases; when combined with CT (PET/CT), it provided both functional and anatomical information, improving diagnostic accuracy [277, 279]. Though the conventional methods, which continue to be used in the hybrid mode (e.g. PET-CT, PET-MRI), remain important for the detection of cancer, they present some limitations in terms of sensitivity and time of detection, delineating true metastatic sites from benign or inflammatory changes and the occurrence of false positives and negatives making the diagnosis of certain cancers quite difficult [276, 277]. Recent advancements in optical imaging have enabled the detection of molecular changes associated with cancer, with techniques such as fluorescence and Raman spectroscopy offering high sensitivity and specificity, facilitating early tumor detection and real-time guidance during surgeries [282]. Figure 3 illustrates the integration of diverse imaging modalities with recent technological advancements to enhance the detection, characterization, and monitoring of cancer metastasis thereby improving diagnostic accuracy and treatment outcomes.

Fig. 3.

Diagnostic tools for detection of metastasis. The figure represents for detection of metastasis (a) Imaging techniques which include ultrasound, bone scintigraphy, single-photon emission computed tomography (SPECT), computed tomography (CT), positron emission tomography (PET) and magnetic resonance imaging (MRI) (b) Biomarkers such as genetic-based markers and blood-based markers, (c) AI and deep learning tools such as AI-assisted metastasis detection, multi-omics based classification and AI-assisted mammograms, (d) Pathological and histopathological tools such as immunohistochemistry, liquid biopsy and tissue biopsy. [Created in https://BioRender.com]

On the other hand, AI techniques, including machine learning and deep learning, are being utilized to predict and manage cancer metastasis by analyzing various molecular data to construct predictive models. However, challenges persist in improving the accuracy and generalizability of these models [283]. Collectively, these progressive developments in imaging technology have significantly enhanced our ability to detect, diagnose, and monitor cancer, leading to improved patient outcomes [277]. Despite the availability of various imaging modalities, tissue biopsy remains the gold standard for cancer detection [276, 284, 285]. It is an invasive procedure where a tissue or cell sample is collected using a needle, circular blade, or surgery, often guided by ultrasonography or CT scan. While it provides high accuracy and stability, it is costly, painful, time-consuming, and carries risks, with limited repeatability for certain tumors and offering only single-point data. Liquid and urine-based biopsies are non-invasive alternatives [276, 285, 286]. A summary of the classical to advanced technologies in studying, detecting and managing various types of metastatic cancer have been described in Table 3.

Table 3.

Classical to advanced imaging systems in studying, detecting and managing various types of cancer and metastasis

Diagnostic procedure	Details of the technique and outcomes	Types of cancer/malignant tumors commonly detected and studied	Advantages	Limitations	References
X-ray imaging (plain)	Low doses of ionizing radiation used to create 2D images of specific areas under investigation	Bones and lung	Simple, cost-effective non-invasive method that provides quick preliminary diagnostic results, portable and convenient	Prolonged use may increase risk of cancer, unsafe for pregnant women, limited image quality, soft tissues do not show well	[287, 288]
Mammograms	Involves X-ray of the breast tissue essentially for early detection and screening of probable malignancy	Breast	Early detection of tumors or irregularities in breast tissue, non-invasive, cost-effective	False positives and false negatives are common place, limitations in detecting subtle lesions	[281, 289]
CT scan (including Spectral CT)	Uses a combination of X-rays and computers to generate 3D cross-sectional images to locate tumors, estimate their size and look for metastasis study recurrence.	Tumor detection size estimation metastatic evaluation and monitoring of recurrence of Bladder, Bowel, Kidney, Ovarian, Stomach, head and neck cancers, lung, breast, pancreatic and prostate cancer.	Quick, non-invasive, detailed images of bones and soft tissue, helps diagnose cancers, blood clots and internal injuries.	Prolonged exposure may increase risk of cancer, contrast dye may not be suitable for patients with kidney ailments, results may be misinterpreted	[277, 290, 291]
Ultrasonography	Uses ultrasonic waves to produce both 2D and 3D images and are useful in detecting superficial cancers in the body	Thyroid nodules, some forms of breast, ovarian, prostate and liver cancers, for guiding tissue biopsies	Non-invasive, uses non-ionizing radiation, hence safe even for pregnant women, portable and can be used at bedside	Cannot penetrate bones or body parts hidden behind bones, disrupted by gas, may lead to false positives	[292]
MRI (including the contrast-enhanced one)	Involves a combination of strong magnets and radio waves to generate 3D computerized images. A intravenously injected contrast dye is used to detect certain malignant tumors in brain and spinal cord	Brain, lung, prostate, liver, bone and soft tissue, spinal cord and breast	Non-invasive, detailed images that provide distinction between fat, muscle and water, covers large areas	Expensive, may affect metal implants like pacemakers, contrast agent may cause allergic reactions in patients, uncomfortable procedure	[281, 293]
Breast MRI	The MRI machine has a special device called a dedicated breast coil to image the breasts.	Certain types breast cancers	High sensitivity, non-ionizing radiation, facilitates early cancer detection	Expensive, may affect implants, contrast agent may cause adverse effects, uncomfortable in some cases, may lead to false positives	[281, 294]
Hyperpolarized [carbon 13 (13C) pyruvate MRI]	Hyperpolarized carbon 13 pyruvate agent is injected into the body, MRI detects exchange of 13C-pyruvate to 13C-lactate over time to asses rate of metabolism	Breast, prostate, liver and brain tumors	Non-ionizing radiation, high sensitivity, early cancer detection and specific to tumor metabolism, metabolic profiling of tumors	High cost, short imaging window, technically complex and requires specialized expertise	[295–297]
MRS (in vivo proton (1H)	Involves the principles of spectroscopy to detect abnormal metabolism in cancer tissues. Helps in cancer prognosis and monitoring treatment responses	Breast and pancreatic cancers	Provides non-invasive measurement of metabolite concentrations, detects changes in steady-state levels of cellular metabolites, and hyperpolarized C MRS enables real-time imaging of metabolic fluxes	Different pathologies’ spectroscopic appearances can overlap, sometimes requiring follow-up or biopsy for definitive diagnosis	[298–301]
SPECT	Involves intravenous injection of a small radioactive tracer (e.g., technetium-99 m, iodine-123) that accumulates in a tumor and a special camera (e.g. gamma camera, rectilinear scanner) takes images of that affected zone. Detecting cancer recurrence and assessing the effectiveness of treatment	Brain and bone cancers and in some cases liver and lung cancer	Relatively safe, SPECT/CT provides improved specificity and accuracy.	Can’t be used on pregnant and breastfeeding individuals, lower resolution and sensitivity when used alone	[302]
PET	Usually use a form of radioactive sugar ([¹⁸F]Fluorodeoxyglucose) that is taken up in large amounts by rapidly growing tumors compared to normal cells. The collected tracer emits gamma rays that are detected by a special camera. In many cases PET is used in combination with CT or MRI to detect malignant tissues. Other radioisotopes are also used based on diagnostic needs	Breast, thryoid, lung, colorectal, esophageal, prostate cancers and melanoma. In some cases, brain tumors and blood cancers	Whole body imaging, PET/CT improves diagnosis, staging and treatment monitoring	Not target specific, Low sensitivity in detection of small tumors, expensive than SPECT, challenges in coronary arteries imaging	[303–305]
Bone and Thyroid scans, Multigated acquisition and Gallium scans (MUGA)	These procedures along with PET, PET/CT are collectively called nuclear medicine scans. Mainly based on body chemistry and involves the use of radiotracers as explained above. These methods help to detect the extent of metastasis and also assess if the treatment is working.	Bone, thyroid. MUGA scans look for heart functions during chemotherapy, whereas gallium scans are used to diagnose lymphoma, leukemia, lung cancer, melanoma, and sarcoma	Provides optimal diagnostic accuracy and treatment in combination with CT or MRI	Limited spatial resolution, multiple acquisition sessions are needed to acquire images and reduced sensitivity to small lesions	[306, 307]
Colonoscopy	Examining the colon and rectum with a flexible tube equipped with a camera.	Colorectal cancer (CRC).	Recommended as the primary minimally invasive examination for individuals at high risk of CRC due to its ability to detect polyps and tumors. Enables tissue biopsy for further analysis. Self-propelling robotic colonoscopes offer painless, adaptable, and dexterous colon examination	Conventional colonoscopy causes pain and discomfort, bowel preparation required, low occurrence of hemorrhage stigmata	[308, 309]
Endoscopy	A flexible tube with a camera is inserted is inserted orally or rectally or both to examine the digestive system	Oesophageal or stomach cancer	Could be used to detect early precancerous lesions, gastric and esophageal neoplasia. Advanced endoscopy improves histopathologic diagnosis. AI with magnifying endoscopy enhances gastric cancer diagnosis	Small lesions can be missed, benign conditions may be misinterpreted, dependent on the expertise of the operator, and depth assessment can be limiting	[310–312]

The transformative role of AI and bioinformatics in cancer diagnosis and metastasis detection: technological advancements and ethical concerns

The development of AI-powered metastasis detection tools marks a transformative shift in cancer diagnosis and treatment by significantly enhancing precision, efficiency, and accuracy. These tools address critical diagnostic challenges, especially in identifying metastases in life-threatening cancers such as brain, lung, and breast. Companies are increasingly leveraging AI technologies across imaging modalities like MRI, CT, EEG, and X-ray to improve diagnostic performance.

AI systems for medical image analysis primarily use advanced deep learning techniques, especially convolutional neural networks (CNNs), which excel in pattern recognition tasks such as tumor detection, lesion segmentation, and anomaly recognition [313]. Variants of CNNs—including DCNNs, ANN, Ksvm, Inception V3, Inception-ResNet V2, ResNet-101, GoogLeNet, 3D CNN, and Faster R-CNN—have been optimized for handling complex tasks, such as analyzing 3D imaging data and detecting subtle abnormalities. Recurrent neural networks (RNNs), though useful for time-series data, are less effective for high-resolution medical imaging due to sequential processing limitations. More recently, transformer-based models like Vision Transformers (ViTs), Data Efficient Image Transformer (DEiT), and BERT Pre-training Image Transformer (BEiT) have emerged for capturing long-range spatial dependencies in medical images [314]. These are often used in encoder-decoder frameworks, where visual features are extracted and translated into text embeddings decoded by models like GPT-2 to generate comprehensive medical reports, improving diagnostic accuracy and reporting efficiency [315].

Commercial tools such as Vysioneer’s FDA-approved VBrain automate tumor contouring for brain metastases, meningiomas, and acoustic neuromas, reducing manual effort and improving radiotherapy planning accuracy [316]. Integrated into clinical workflows, such tools minimize diagnostic errors and improve real-time decision-making. Similarly, companies like Philips (IntelliSpace AI) [317] and Behold.ai (Red Dot) [318] have secured patents and regulatory approvals like FDA clearance and "Conformité Européenne" (CE) marking, boosting clinical adoption and investor confidence.

AI-powered tools such as Northwell Health’s iNav for pancreatic cancer [319] and Turing.com’s melanoma detection system [320] enable early identification of imaging anomalies often missed in conventional screenings. These systems process vast datasets rapidly, assisting radiologists in detecting microscopic metastases and enhancing diagnostic accuracy, especially in resource-limited settings where access to expert radiologists is limited.

Nevertheless, these systems require high-quality, labeled datasets and significant computational resources. Concerns regarding bias in training data, data privacy, and lack of interpretability remain barriers to broader clinical acceptance [321].

Companies like Google Health [322] and Paige [323] are transforming pathology by using AI to detect metastatic breast cancer in lymph nodes, reducing unnecessary biopsies and enabling timely treatment. Tools from PathAI and Proscia demonstrate how deep learning models can enhance diagnostic precision by identifying subtle histopathological patterns that may elude human observation. PathAI’s offerings—AIM-PD-L1 (NSCLC), AIM-HER2 (breast cancer), and ArtifactDetect—improve diagnostic reliability [324, 325]. Proscia’s Concentriq^® platform integrates AI-driven workflows and embeddable tools to streamline data analysis and reduce diagnostic turnaround time [326].

Beyond diagnosis, AI enables precision oncology by identifying genetic mutations relevant to drug response. Recursion Pharmaceuticals and Numerate use AI for drug discovery and therapeutic prediction. Recursion’s ecosystem—including Recursion OS, OpenPhenom-S/16, BioHive-2 supercomputer, and the custom large language model (LLM) agent LOWE—integrates imaging, high-throughput screening, and generative AI to accelerate the discovery of effective, individualized cancer treatments [327, 328].

Despite these advancements, concerns remain regarding over-reliance on AI, potentially diminishing the role of human expertise in cancer diagnostics. However, AI is designed to augment rather than replace clinician’s expert assessment. Platforms like Roche’s Digital Pathology Open Environment integrate AI algorithms with human expertise, ensuring that AI-assisted findings are validated by medical professionals. Roche’s Digital Pathology Open Environment integrates cutting-edge deep learning algorithms to enhance cancer diagnostics. As part of the navify Digital Pathology platform, it incorporates over 20 AI tools from eight key collaborators—Deep Bio (utilizes supervised learning on high-resolution pathology images for precise tumor segmentation for prostate cancer detection, tumor quantification and grading), DiaDeep (employs feature extraction layers to detect and quantify breast cancer biomarkers in stained tissue samples), Lunit (integrates tumor segmentation and PD-L1 expression quantification, Tumor Proportion Score (TPS) evaluation for NSCLC), Mindpeak (pre-trained on extensive datasets and optimized for precise biomarker detection tasks), Owkin (integrates diverse CNN architectures to enhance accuracy in microsatellite stability detection in CRC), Qritive (integrates hierarchical layers for multi-level tissue sample classification, prostate cancer screening and grading, lymph node metastasis analysis, and colon cancer detection), Sonrai Analytics (microsatellite instability (MSI) determination status in CRC), and Stratipath (risk-profiling of invasive breast cancer)—advancing precision medicine and pathology insights [329]. These models mostly use CNN architectures to analyze and support multi-cancer diagnostics, including prostate, breast, lung, and colorectal cancers, enabling tumor detection, biomarker quantification (e.g., HER2, PD-L1), and microsatellite instability screening (MSS/MSI) [329].

AI’s global scalability is evident in companies like Canon [330] and Visiopharm [331], which have deployed AI-based diagnostics across varied healthcare environments. In low-resource settings, AI tools can democratize access to expert-level diagnostics by automating complex evaluations.

However, critical comparative evaluations of AI tools are often lacking. Factors such as generalizability across populations, clinical robustness, interpretability, and regulatory consistency are not uniformly addressed, limiting clear assessment of clinical utility and reliability as discussed in Table 4.

Table 4.

Comparative study of AI-powered metastasis detection tools (with accuracy metrics and limitations)

Tool/company	Modality	Model type	Strengths	Limitations	Regulatory status
Google LYNA	Histopathology (lymph nodes)	Deep CNN	High sensitivity for breast cancer micrometastasis	Generalizability to other cancers limited	FDA cleared (research use)
Vysioneer VBrain	MRI (brain metastasis)	Deep Learning CNN	Automates tumor contouring, speeds radiotherapy planning	Limited to specific brain tumors	FDA approved
PathAI	Histopathology (various)	Ensemble CNN and SlideQC tools	Integrates PD-L1/HER2 scoring; slide quality analysis	Interpretability challenges; limited transparency	Used in trials/validation
Paige.AI	Prostate, breast pathology	CNN-based and Transformer variants	Strong performance in prostate cancer detection	Data dependency, model bias risks	FDA approved
Proscia (Concentriq®)	Digital pathology (various)	CNN and AI Toolkits	Workflow integration; supports AI embedding and plugin models	Needs large computational setup	CE marked; limited FDA use
Philips IntelliSpace AI	Radiology (general)	Deep Learning (various)	Modular, scalable platform with integration options	Some models lack clinical validation	CE marked
Behold.ai Red Dot	X-ray (chest)	AI-Deep Learning	Emergency room triage for pneumothorax and lung masses	Performance may degrade in low-quality imaging	FDA cleared
Recursion Pharma	Drug discovery and imaging	Deep Learning and LLMs	Identifies therapeutic targets using imaging-genomics integration	Not focused on diagnostics; very data-hungry	Not diagnostic FDA scope
Roche navify® Digital Pathology	Multi-cancer pathology	CNN (via collaborators)	Combines > 20 AI tools for tumor typing, grading, MSI analysis	Complexity and cost of platform integration	CE marked; clinical trials
Canon and visiopharm	Various imaging modalities	CNN and ML models	Scalable AI applications in diverse regions	Infrastructure and training costs	CE marked

AI’s integration with high-throughput sequencing and metagenomics is enabling microbiome-informed personalized treatments. Algorithms like Random Forest and artificial neural networks are being used to classify cancer types based on tissue-specific microbial profiles and identify viral sequences in metagenomic data [264, 332].

While some models have obtained FDA or CE approval, many still face significant regulatory and ethical challenges due to the lack of rigorous oversight. Ethical issues—including algorithmic bias, data privacy, and liability—hinder the adoption of these detection tools. A 2022 Japanese study reported an 11% diagnostic error rate in AI-assisted outpatient care, with errors more frequent when correct diagnoses were absent from AI suggestions [333]. In a 2023 randomized clinical vignette study, clinicians relying on biased AI models had an 11.3-point drop in diagnostic accuracy, with specificity falling from 73 to 53%, and image-based AI explanations failing to correct these errors [334]. Additionally, certain AI models demonstrated 22% lower accuracy in detecting sclerotic versus lytic bone metastases due to training data imbalances [335].

These findings underscore systemic risks such as automation bias, inadequate oversight, and training data limitations. They highlight the need for robust human-AI collaboration, continuous performance audits, and validation tailored to clinical contexts.

While AI-powered metastasis detection tools offer significant advancements, their widespread adoption hinges on addressing challenges like data quality, bias mitigation, interpretability, regulatory validation, and clinician training. Platforms like Qure.ai and Philips are at the forefront of multi-modal diagnostics, but sustained collaboration among developers, clinicians, and regulators is essential to ensure these technologies are accurate, ethical, and accessible—ultimately enhancing patient outcomes and driving forward the paradigm of precision oncology.

Conclusion

Addressing the global cancer burden necessitates a multifaceted approach that integrates environmental, microbial, and genomic influences on metastasis with cutting-edge technological advancements in detection and treatment. This review uniquely synthesizes the impact of diverse external factors on metastatic progression with emerging technologies, offering a comprehensive perspective on metastasis research. Pollution, chemical exposures, and microbiome dysbiosis significantly contribute to metastatic progression, while organotropism and genomic variations dictate metastatic patterns. Advanced detection tools, including in vitro models, liquid biopsies, genome sequencing, and AI-driven imaging, have revolutionized metastasis research, enabling early diagnosis and personalized treatment strategies. The emergence of CRISPR-Cas9, metabolomics, and single-cell transcriptomics further refines our understanding of metastatic mechanisms. However, significant challenges remain, which includes the following:

• Inadequate integration of environmental carcinogens in metastasis research.

• High inter-patient heterogeneity in metastatic disease.

• Limited clinical translation and standardization of detection methods of emerging diagnostic and therapeutic technologies.

• Regulatory hurdles in novel therapeutics and disparities in accessibility of advanced technologies in resource-limited settings.

• Ethical concerns surrounding the use of AI in cancer diagnostics.

Future research must focus on improving the clinical translation of emerging technologies, developing precision medicine approaches tailored to metastatic heterogeneity, and enhancing predictive models for therapy response; however, addressing metastatic cancer effectively requires more than scientific breakthroughs—it demands robust policy reforms to overcome barriers in access, implementation, and ethical oversight. Some key areas which may be addressed on priority are stated below.

1. Cutting-edge technologies like liquid biopsies, CRISPR-based therapies, AI diagnostics, and single-cell omics remain confined to research or elite clinical settings. Without supportive policy frameworks, these innovations cannot be translated into standard practice or scaled across diverse health systems. Reforms are needed to:

   • Streamline clinical validation and approval processes.
   • Create incentives for public-private translational partnerships.
   • Integrate novel tools into national cancer care guidelines.
2. High costs, lack of infrastructure, and workforce shortages make advanced cancer care inaccessible in many low- and middle-income countries (LMICs). Policy interventions are needed to:

   • Subsidize or tier pricing of diagnostics and therapies.
   • Support cross-border technology transfer.
   • Build local capacity through workforce training and infrastructure development.
3. The rapid rise of AI and data-driven technologies introduces risks of bias, privacy breaches, and misuse, especially in vulnerable populations. Current regulations are often outdated or non-existent for such technologies. Reforms must:

   • Establish clear ethical frameworks for AI use in oncology.
   • Mandate transparency, algorithmic fairness, and informed consent.
   • Harmonize regulatory processes for international collaboration and data sharing.
4. Despite growing evidence linking environmental exposures (pollution, chemicals, microbiome disruption) to cancer metastasis, most cancer control policies overlook environmental determinants. Reforms are required to:

   • Expand environmental carcinogen monitoring and exposure limits.
   • Integrate environmental health with cancer surveillance programs.
   • Develop regulations that reduce exposure to known metastatic triggers.

5. Current cancer policies are often reactive rather than predictive. Integrating AI-powered risk modeling using environmental, genomic, and clinical data could revolutionize early intervention. However, without supportive data policies and funding, such systems cannot be effectively developed or deployed.

A comprehensive and equitable approach to metastatic cancer care requires aligning scientific innovation with ethical regulation, environmental accountability, and global accessibility. Policies grounded in this multidimensional framework will not only improve survival outcomes but also promote health equity across nations.

Abbreviations

2D and 3D

2-dimensional and 3-dimensional

ALK

Anaplastic lymphoma kinase

AR

Androgen receptor

BCR-ABL

Breakpoint cluster region–Abelson murine leukemia viral oncogene homolog 1

BMP

Bone morphogenetic protein

BRAF

B-type rapidly accelerated fibrosarcoma (part of the RAF kinase family) 

Cas9

CRISPR-associated protein 9

CCNH-C5orf30

Cyclin H (“CCNH”) and chromosome 5 open reading frame 30 (“C5orf30”) genes

CD

Cluster of differentiation

CDH1

Cadherin 1 (also known as epithelial cadherin or E-cadherin)

CDKN2A

Cyclin dependent kinase inhibitor 2 A encoding gene

cfDNA

Cell-free DNA

CIN

Chromosomal instability

CRC

Colorectal cancer

Cre–LoxP

Causes recombination– Locus of crossover (x) in P1 bacteriophage (Cre refers to Cre recombinase and LoxP is a 34-base pair DNA sequence recognized by Cre recombinase)

CRISPR

Clustered regularly interspaced short palindromic repeats

CTCs

Circulating tumor cells

ctDNA

Circulating tumor DNA

CTLA-4

Cytotoxic T-lymphocyte antigen-4

CXCL12

C-X-C motif chemokine ligand 12

CXCR4

C-X-C motif chemokine receptor 4

ECM

Extracellular matrix

EGFR

Epidermal growth factor receptor

EMT

Epithelial-to-mesenchymal transition

ERK

Extracellular regulated kinase

EVs

Extracellular vesicles

FDA

The United States Food and Drug Administration

HCC

Hepatocellular carcinoma

HDI

Human development index

HER2

Human epidermal growth factor receptor 2

HNSCC

Head and neck squamous cell carcinoma

IL6

Interleukin 6

IVM

Intravital microscopy

KRAS

Kirsten rat sarcoma viral oncogene homolog

LB

Liquid biopsy

macsGESTALT

multiplexed, activatable, clonal and subclonal genome editing of synthetic target arrays for lineage tracing

MDSCs

Myeloid-derived suppressor cells

MET

Mesenchymal-to-epitheliall transition

MIR

Mortality-to-incidence ratio

MMPs

Matrix metalloproteinases

MRS

Magnetic resonance spectroscopy

MUC16

Mucin 16

NGS

Next generation sequencing

NSCLC

Non-small cell lung cancer

PCR

Polymerase chain reaction

PD-L1

Programmed cell death ligand 1

PDX

Patient-derived xenografts

PDX1

Pancreatic and duodenal homeobox 1 encoding gene

PIK3CA

Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha encoding gene

PM

Particulate matter

PTEN

Phosphatase and tensin homolog encoding gene

RANKL

Receptor activator of nuclear factor κB ligand

RAS

Rat sarcoma virus oncogene

RB1

Retinoblastoma 1 encoding gene. The mouse ortholog of the human RB1 gene is referred to as Rb

RET

REarranged during transfection

RNAi

RNA interference

ROS1

ROS protooncogene 1, encoding a receptor tyrosine kinase

SCFAs

Short-chain fatty acids

SNAIL

Snail family transcriptional repressor

TGF-β

Transforming growth factor-beta

TME

Tumor microenvironment

TP53/Trp53

Tumor protein p53 encoding gene (in human beings)/Transformation related protein (in mice). The Trp53 gene in mice is orthologous to the human TP53 gene

TRAPs

Traffic-related air pollutants

Tregs

Regulatory T cells

TRMT11-GRIK2

tRNA methyltransferase 11 homolog (“TRMT11”) and glutamate receptor, ionotropic, kainate 2 (“GRIK2”) genes

TWIST1

Twist family BHLH transcription factor 1

VEGF

Vascular endothelial growth factor

WGD

Whole-genome doubling

WNT

Wingless/Integrated. The term "WNT/Wnt" is a portmanteau of wingless (a Drosophila segment polarity gene) and Int-1 (a gene into which the mouse mammary tumor virus (MMTV) integrated to cause tumors)

YAP

Yes-associated protein

TAZ

Transcriptional coactivator with PDZ-binding motif

Author contributions

S.B.—Conceptualization, visualization, data curation, formal analysis, writing original draft, review, editing, supervision and funding acquisition; S.Sasikumar—formal analysis, data curation, writing original draft, review, and editing. S.Sur—review and editing, resources, and funding acquisition; V.V.— data curation, formal analysis, writing original draft; S.K.—data curation, formal analysis, writing original draft; S.G.— data curation, formal analysis, writing original draft; A.R.— data curation, formal analysis, writing original draft; M.P.— data curation, formal analysis, writing original draft; N.N.— data curation, formal analysis, writing original draft; A.B.— data curation, image creation; N.A.— data curation, image creation. S.B. and S. Sasikumar contributed equally.

Funding

This work is supported by the Ramalingaswami Re-entry fellowship, Department of Biotechnology, Govt. of India to S. Sur [BT/RLF/Re-entry/47/2021] and Intramural Grants, Dr. D. Y. Patil Vidyapeeth (DPU), Pimpri, Pune, India to S. Basu [DPU/644 − 43/2021] and, [DPU/1211 (3) /2024], and to S. Sasikumar [DPU/1211 (11) /2024] and to A Ranjan [DPU/1211 (2) /2024].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.