Advancing prussian blue nanoparticle-mediated photothermal therapy through machine learning and multiomics integration

ABSTRACT

Prussian blue nanoparticles (PBNPs) are a versatile platform for administering photothermal therapy (PTT) in cancer therapy applications. PBNPs combine biocompatibility, safety, and clinical translational potential with durable treatment outcomes in preclinical cancer models. In this perspective, we focus on aspects critical to the workflow of implementing PBNP-PTT in cancer treatment, drawing inspiration from adjacent scientific areas that have not been described in the context of PBNPs, but are important for improving the delivery of PBNP-PTT and its translation. Specifically, we will discuss machine learning approaches, multiomics analyses, and clinical strategies pertinent to PBNP-PTT. Machine learning approaches have the potential to enhance PBNP-PTT design, performance, and therapeutic outcomes. Complementing this, multiomics has the potential to describe the responses to PBNP-PTT, particularly its immune effects. By embedding these advances from nanoparticle engineering to therapy monitoring, PBNP-PTT can evolve from empirical tumor ablation toward a precision photothermal platform, enabling highly individualized cancer treatments with improved safety, efficacy, regulatory approval, and clinical predictability. We will also cover clinical strategies pertinent to the translation PBNP-PTT culminating with specific forward-looking perspectives. This convergence of nanotechnology, immunology, and data science positions PBNP-PTT at the forefront of next-generation cancer nanomedicine and immunotherapy.

1. Introduction

Nanoparticle-mediated photothermal therapy (PTT) has emerged as a powerful experimental cancer therapy. PTT exerts its antitumor effects as a localized thermal therapy that disrupts tumor cell integrity, activates immune responses and improves immune cell infiltration within the tumor microenvironment (TME) [1]. PTT has demonstrated synergy with immunotherapy in diverse preclinical cancer models, exerting durable and persistent treatment outcomes [2]. In particular, Prussian blue nanoparticles (PBNPs) have gained particular attention as PTT agents (henceforth PBNP-PTT) due to their excellent biocompatibility, well-characterized safety profile, including FDA approval for clinical use in heavy metal poisoning, high photothermal conversion efficiency, and multifunctional potential [2]. Upon accumulation or direct injection of PBNPs within the TME and subsequent irradiation with a near-infrared (NIR) light source, typically a laser, PBNPs convert light energy into heat (Figure 1). This thermal stress induces cell death and promotes immunogenic cell death (ICD), characterized by the release of tumor-associated antigens and danger signals, including damage associated molecular patterns (DAMPs), that enhance antigen presentation and immune cell activation (Figure 1) [3,4]. The ability to combine thermal ablation with immune activation makes PBNP-PTT a promising platform in cancer immunotherapy applications transforming localized tumor destruction into a systemic immune response that enhances the overall treatment outcomes, facilitating the regression of distant lesions through an abscopal effect [5,6].

Figure 1.

Envisioned ML – multiomics framework for optimizing Prussian blue nanoparticle – mediated photothermal therapy (PBNP-PTT) as part of a synergistic closed-loop. ML frameworks can incorporate diverse variables such as PBNP design, treatment parameter optimization, laser modality, and immune response prediction to improve the delivery of PBNP-PTT. Within a closed-loop framework, these ML models can iteratively learn from experimental outcomes and feed optimized design and treatment parameters back into subsequent PBNP-PTT cycles. These approaches can facilitate improved tumor ablation and concomitant favorable antitumor immune effects. Simultaneously, multiomics platforms, including transcriptomics, proteomics, metabolomics and lipidomics, can generate comprehensive profiling of responses to PBNP-PTT. These high-dimensional multiomics datasets provide system-level biological feedback that informs ML model training and refinement as well as better mechanistic understanding of the effects of PBNP-PTT. Together, ML and multiomics operate synergistically as an iterative optimization loop, linking nanoparticle design, treatment parameters, and biological response. This integrated workflow can facilitate the clinical translation of PBNP-PTT as well as the design of new combination treatment strategies using PBNP-PTT. Created in BioRender. Sweeney, L. (2026) https://BioRender.com/o6gqbuf.

Despite strong preclinical efficacy, the clinical translation of PBNP-PTT remains limited by an incomplete understanding of how to rationally optimize nanoparticle design, treatment parameters, and immune engagement across heterogeneous tumors and patient populations. Currently, PBNP-PTT studies largely rely on empirical formulations and fixed laser protocols, with limited integration of data-driven optimization or systems-level immune profiling. Additionally, standardized strategies for image-guided delivery, treatment monitoring, and regulatory alignment remain underdeveloped. As the field moves toward precision cancer nanomedicine, there is a critical need for approaches that integrate machine learning, multiomics immune profiling, and translational workflow design to enable reproducible, personalized, and clinically scalable PBNP-PTT. Therefore, addressing these gaps is essential to clinically advancing PBNP-PTT.

In this perspective, we highlight four key areas within the PBNP-PTT workflow that remain underexplored: machine learning approaches, multiomics analyses, clinical translation strategies, and forward-looking directions for the field. To address these gaps, we draw inspiration from studies and reports in PTT-adjacent nanosystems, as limited work has specifically explored these aspects for PBNPs. We begin by discussing the potential of machine learning to enhance PTT performance and predict therapeutic outcomes. Emerging computational machine learning tools hold the potential to accelerate nanoparticle design, optimize treatment parameters, and improve patient selection. We then highlight how multiomics approaches can elucidate the immune consequences of PBNP-PTT. These approaches enable unprecedented insights into tumor – immune interactions, offering opportunities to dissect the molecular consequences of PBNP-PTT and guide the rational design of combinatorial therapies. Next, we address the current status and challenges of clinical translation, drawing parallels with other nanoparticle systems and identifying key regulatory and safety considerations. Finally, we present future directions, envisioning how PBNPs could evolve into versatile platforms that integrate therapy, imaging, and immune modulation within the broader landscape of cancer nanomedicine.

Collectively, these forward-looking innovations position PBNPs at the intersection of nanotechnology, systems immunology, and data science, establishing a foundation for next-generation precision cancer nanomedicine. By focusing on these emerging and under-investigated areas, this perspective complements existing literature and provides a roadmap for advancing PBNPs from bench to bedside in cancer immunotherapy.

2. Machine learning (ML) approaches to enhance PBNP-PTT

The therapeutic success of PTT relies on precise control over multiple interdependent variables, including nanoparticle design, their biodistribution, laser parameters, and tumor biology. While PBNPs offer high photothermal conversion efficiency and favorable biocompatibility, optimizing their performance across diverse tumor and patient contexts remains challenging. ML provides powerful tools to address this complexity, offering data-driven strategies to accelerate the design, prediction, and personalization of PBNP-PTT (Figure 1).

2.1. Nanoparticle design and property prediction

ML algorithms can be trained on large datasets of nanoparticle physicochemical features including size, shape, surface chemistry, and aggregation state to predict key performance outcomes such as optical absorption, photothermal conversion efficiency, and circulation half-life. Recent studies have already begun applying ML-assisted modeling to nanoparticles to relate their structural features to optical and photothermal behavior [7,8], but substantial opportunity remains for systemic across design landscapes. For PBNPs specifically, the lattice composition is fixed (Fe²⁺/Fe³⁺ –CN – Fe framework); significant changes in photothermal conversion have been shown to generally require compositional modification such as ion doping or ion substitution within the cyanometallate lattice [9]. ML could systematically evaluate how compositional modifications such as metal ion substitution, lattice doping or hybrid architectures affect photothermal behavior to optimize therapeutic performance of PBNPs.

Direct ML integration with PBNP-based materials remains limited. Hof et al. used AI assisted continuous flow synthesis with Bayesian optimization and Gaussian process regression to control particle size, morphology, and compositional homogeneity but focused on nanoparticle synthesis rather than functional optimization [10]. In contrast, ML-driven compositional optimization has been to shown to bolster the discovery of other high performance photothermal nanomaterials. For example, Fan et al. used XGBoost and feature engineering to link composition to photothermal conversion efficiency in metal – phenolic networks (MPNs) and identify iron based systems with exceptional heating performance [8]. Although MPNs and PBNPs differ in their compositional design space, the ML paradigm utilized by Fan et al. can be adapted to PBNPs by transferring regression models, feature engineering strategies and virtual screening workflows that link composition to photothermal output and reduce experimental burden. By redefining model inputs to include PBNP specific descriptors such as certain lattice metal substitutions (e.g., Mn²⁺, Cu²⁺, Co²⁺) that affect photothermal efficiency by modifying intervalence charge-transfer pathways and the electronic band structure, ML frameworks developed for other photothermal nanomaterials can be systematically tuned to optimize PBNP performance [11,12].

Building on these material-specific ML adaptations, studies emphasizing model interpretability provide further insight into photothermal materials design. Wu et al. used Random Forest regression models with SHapley Additive exPlanations (SHAP) analysis to identify molecular features with high photothermal conversion in organic agents, such as quaternary ammonium groups, methoxy groups and tetraphenylethylene moieties, and experimentally validated these predictions [13]. Although this chemical composition optimization approach may not be directly applicable to fixed lattice PBNPs, the principle of linking structure to photothermal performance is highly relevant. Similar ML frameworks could identify lattice substitutions, mixed-metal compositions and structural parameters (e.g., size, morphology, porosity, aggregation) that most effectively enhance PBNP photothermal behavior

2.2. Optimization of treatment parameters

The therapeutic index of PTT is determined not only by nanoparticle properties but also by treatment parameters such as laser wavelength, power density, exposure duration and nanoparticle dosing. ML-driven optimization frameworks can model the interaction of these variables to maximize tumor ablation while minimizing off-target heating and tissue injury. Coupled with real-time feedback systems such as thermal imaging or photoacoustic monitoring, adaptive AI systems may enable individualized control of PTT during treatment sessions.

Varon et al. integrated NIR laser intensity and exposure duration with cytotoxicity data to develop ML models predicting treatment efficacy for photodynamic therapy (PDT), PTT, and their combination [14]. Using gold nanoparticles conjugated with photosensitizers in neuroblastoma cells, they tested various laser intensity and exposure duration parameters on neuroblastoma cells in-vitro and employed analytical fitting to map complex parameter interactions. The model not only predicted cell death probability as a continuous function of laser intensity and duration but also quantified synergistic gains when combining PDT and PTT. While Varon et al. focused on in vitro tumor cell killing, the underlying ML framework is readily transferable to PBNP-PTT by redefining model outputs to include immunogenic endpoints such as ICD markers, cytokine release profiles, or immune cell recruitment as target variables, allowing ML models to identify laser – nanoparticle combinations that maximize immune productive damage rather than bulk cytotoxicity. Specifically, training datasets could incorporate PBNP characteristics (size, shape, surface coating, photothermal conversion efficiency, optical absorption spectrum), treatment parameters (laser power, duration, wavelength, spot size, nanoparticle dose, tumor depth), and biological response markers of interest such as ICD indicators (calreticulin exposure, HMGB1 release, ATP secretion) as well as other immune- or tumor-related readouts. ML optimization could then identify PBNP and laser combinations that maximize therapeutic benefit while minimizing normal tissue damage, advancing personalized immunomodulatory PTT protocols.

Shirisha et al. further demonstrated this concept by developing a convolutional neural network (CNN) to predict optimized thermal doses in nanoparticle-mediated PTT [15]. Using a synthetic dataset spanning over 10,000 samples with varying nanoparticle concentrations, laser intensities, exposure times, and tissue absorption coefficients, their model achieved high testing accuracy [15]. For PBNP-PTT, this CNN based framework could be adapted to incorporate PBNP-specific optical and thermal descriptors, such as NIR absorption strength, lattice-dependent heat generation efficiency, and thermal relaxation times. Additionally, the model’s strong generalizability suggests potential for tailoring PBNP-PTT protocols to individual patient profiles, accounting for tumor type, microenvironment characteristics, and patient-specific factors. Future iterations could integrate additional inputs such as immunogenicity profiles, hypoxia levels, immune cell infiltration patterns, and patient genomic data to predict not only thermal dose but also immune activation markers, T cell expansion, and the likelihood of durable tumor regression or relapse, building on literature precedent with these approaches [16,17].

2.3. ML-Driven pulsatile PTT

The mode of laser irradiation critically affects both tumor ablation mechanisms and immune response, with important implications for PBNP-PTT. Continuous wave (CW) lasers deliver steady light flux over time, producing gradual heating that typically raises local temperatures to 42–50°C, sufficient to induce protein denaturation, membrane disruption and apoptosis or necrosis. Huang et al. demonstrated that under CW irradiation, apoptosis occurs at lower energies for cytoplasm-targeted gold nanoparticles, while necrosis requires higher doses (108–114 J) and nucleus-targeted particles rarely reach necrosis thresholds even at 210 J [18]. This predictable thermal profile makes CW irradiation suitable for bulk tumor ablation but the broad zones of heating can cause collateral tissue damage [19].

In contrast, pulsed lasers deliver high-intensity energy in discrete bursts, generating rapid, localized hyperthermia. Huang et al. found that nanosecond-pulsed lasers induced immediate necrosis regardless of nanoparticle localization, with nucleus-targeted gold nanoparticles requiring low energy thresholds (0.3–0.45 mJ per pulse) due to enhanced plasmonic heating from nanoparticle aggregation [18]. Farivar et al. extended this in vivo, demonstrating that long-pulsed PTT (3–100 ms) with gold nanorods produced confined lesions whereas CW irradiation at equivalent total energy caused wider necrosis (3–6 mm), skin blistering and off-target heating [20]. Pulsed PTT also enhanced immunogenic effects in mice with bladder tumors, increasing ATP release, T cell cytotoxicity, proinflammatory cytokines, and dendritic cell maturation in tumor-draining lymph nodes [20]. When combined with anti-PD-L1 checkpoint blockade, pulsed PTT delayed tumor growth and improved survival, whereas anti-PD-L1 alone was ineffective [20]. It has not yet been evaluated whether PBNPs exhibit comparable advantages under pulsed irradiation, specifically whether their heat generation mechanisms and thermal relaxation timescales support nanosecond or millisecond pulsing. This represents a knowledge gap in the field. Consequently, dedicated experimental studies probing PBNP thermal relaxation kinetics across pulse widths are therefore required to determine optimal irradiation regimes.

Recent ML-guided studies further highlight pulsed PTT’s precision. Kanelli et al. used poly(caprolactone) microparticles loaded with MoS₂ nanosheets under pulsatile NIR laser irradiation, optimizing laser power, pulse duration, cycle number and drug loading via a Classification and Regression Tree (CART) model [21]. Three cycle pulsatile irradiation achieved complete tumor regression and doubled survival without off-target damage, demonstrating precise thermal and drug release control [21]. For PBNP- PTT, these findings suggest that pulsed irradiation could offer spatially selective ablation and enhanced immunogenic cell death, minimizing collaterally injury. CW irradiation may be preferred for tumors needing wider ablation margins or less sensitive anatomical sites, whereas pulsed regimes may prove advantageous when precise thermal control is necessary, although these potential advantages are yet to be tested for PBNP-PTT.

2.4. Predicting biodistribution and immune responses

Beyond optimizing physical parameters, ML can predict both nanoparticle biodistribution and the immune responses they elicit, enabling enhanced design of immunomodulatory PBNP-PTT. Yu et al. developed a Tree-Based Random Forest Feature Importance and Interaction Network Analysis (TBRFA) framework to interpret complex datasets linking nanoparticle properties to biological outcomes [22]. Using lung tumor samples from mice and rats, the model incorporated features spanning composition, size, surface, chemistry, animal models and experimental conditions [22]. TBRFA assessed feature importance and constructed interaction networks that revealed how nanoparticle characteristics jointly influence pulmonary immune responses and organ accumulation and achieved high predictive accuracy for cytokine responses and nanoparticle lung burdens [22]. Applied to PBNP-PTT, this framework could be adapted to identify which PBNP physicochemical features most strongly govern tumor accumulation and immune activation, to inform optimization of particle size, surface chemistry and dosing strategies.

Complementing biodistribution prediction, Becharef et al. developed a high-throughput, ML-assisted assay to evaluate macrophage responses to nanoparticles and PTT, specifically targeting the reprogramming of immunosuppressive M2 macrophages into proinflammatory M1 phenotypes [23]. Using THP-1-derived and primary human macrophages polarized into M0, M1, and M2 states, they exposed cells to various nanoparticles (gold salts, gold-iron oxide nanoflowers, magnetosomes) with and without PTT activation [23]. Immune responses were quantified via gene expression profiling, cytokine secretion assays, and NF-κB/IRF pathway activity measurements and integrated using principal component analysis (PCA) to rank nanoparticle-PTT combinations by immunomodulatory potency [23]. Certain nanoparticles, particularly gold-iron oxide nanoflowers under PTT, efficiently reprogrammed M2 macrophages to M1 phenotypes, enhancing proinflammatory signaling [23]. Validation in primary human macrophages confirmed that the multivariate model captured physiologically relevant immune plasticity, demonstrating the predictive power of ML approaches for designing nanoparticle-mediated immunotherapies [23]. This approach could be directly extended to PBNP-PTT to screen PBNP formulations and irradiation conditions for their ability to reprogram tumor associated macrophages and promote a proinflammatory TME. PBNP was previously shown to reprogram tumor-associated macrophages from pro-tumor M2 to anti-tumor M1 phenotypes in oral squamous cell carcinoma [24].

2.5. Relevance of existing ML frameworks and gaps for PBNP-PTT

The full potential of ML in PBNP-PTT lies in integrating multiple levels of biological complexity into comprehensive predictive frameworks. By integrating physicochemical nanoparticle descriptors, high-throughput immune profiling, and clinical outcome data, ML frameworks could simultaneously predict biodistribution, tumor accumulation, and immune activation. This enables rational optimization of PBNP size, surface functionalization, dosing, and treatment parameters to maximize both tumor ablation and immunomodulatory effects, transforming PBNP-PTT into a data-driven, personalized therapeutic strategy.

However, critical gaps remain for robust ML – PBNP integration. Existing models largely rely on descriptors optimized for more chemically flexible systems than PBNPs and do not capture the lattice dependent electronic and structural determinants of photothermal conversion. Moreover, experimental datasets linking PBNP composition to thermal relaxation kinetics, pulse duration dependent responses, and immunogenic cell death remain limited, constraining model training and validation. Addressing these gaps through PBNP-specific, multimodal datasets and lattice informed ML frameworks that integrate treatment parameters and immune readouts will be essential for translating ML optimized PTT into predictive, patient personalized and immunologically informed PBNP-PTT strategies (Table 1).

Table 1.

Machine learning methodologies applied in related photothermal systems with prospective applicability to PBNP-PTT.

Design or System Variable	ML Approaches Demonstrated in Related Photothermal Systems	Relevance to PBNP-PTT Optimization	References
PBNP physicochemical design (size, coating, optical absorption)	Regression models (e.g., XGBoost, Random Forest, Gaussian process regression) linking design descriptors to desired response; SHAP based interpretability to identify performance driving features	Predict heat conversion efficiency, stability, and tumor accumulation	[8,10,13]
Treatment parameter selection	Multivariate ML models integrating treatment parameters and toxicity data to map relationship between parameter and therapeutic outcome	Personalization of dose – irradiation combinations	[14,21]
Laser delivery mode (CW vs pulsed, fluence, duration)	CNNs and decision tree based models trained on thermal datasets to predict spatial-temporal heating outcomes under different irradiation regimes	Thermal confinement and immunogenic cell death control	[15]
Biodistribution and tumor accumulation	TBRFA linking nanoparticle properties to organ accumulation and tumor uptake	Patient-specific targeting and off-target heating risk	[22]
Immune response to PBNP-PTT	ML-assisted multivariate immune profiling with PCA integrating cytokine secretion, gene expression, and pathway activation data following nanoparticle-mediated PTT	Predict synergy with immunotherapy and abscopal potential	[23]

3. Multiomics integration with PBNP-PTT

While ML provides powerful tools to optimize the delivery of PBNP-PTT, its predictive performance ultimately depends on the quality and biological relevance of input data. Multiomics profiling offers a complementary and essential layer by capturing the molecular and immune programs activated by PBNP-PTT. Integrating omics-derived biomarkers with ML models enables data-driven identification of response signatures, immune activation states, and resistance mechanisms, thereby transforming PBNP-PTT from empirically optimized therapy into a biologically informed, precision-guided intervention. The downstream biological effects of PBNP-PTT are highly complex, involving changes at the transcriptomic, proteomic, metabolomic, and single-cell landscapes (Figure 1) [3,4]. This section highlights multiomics approaches that provide powerful tools to dissect these multilayered responses and guide the rational design of PBNP-PTT.

3.1. Transcriptomic insights into PBNP-PTT

RNA sequencing (RNA-seq) can capture transcriptional programs activated after PTT, including stress responses, immune checkpoint upregulation, and cytokine release. RNA-seq of solid tumors treated with molybdenum-doped PBNP coated with a cancer cell membrane (PMo@CCM) revealed extensive transcriptomic reprogramming linked to improved antitumor immunity [25]. The PMo@CCM nanoplatform integrates photothermal activity with chemodynamic therapy (CDT) via Mo/Fe-mediated glutathione depletion and ROS generation, while the homologous cancer cell membrane coating enhances tumor targeting and antigen supply [25]. Transcriptomic analysis showed upregulation of genes involved in immunogenic cell death, antigen processing and presentation, dendritic cell maturation, and T-cell activation, together with downregulation of tumor growth and immunosuppressive pathways [25]. These RNA-seq findings mechanistically support the observed remodeling of the tumor immune microenvironment, enhanced synergy with PD-1 checkpoint blockade, and robust inhibition of primary and metastatic solid tumors. Cuproptosis is copper-induced regulated cell death causing mitochondrial stress [26]. A Cu²⁺ -doped hollow Prussian blue nanoplatform (Cu-HPB) loaded with cholesterol oxidase was shown to effectively deplete cholesterol in breast cancer cells and inhibit metastasis, with transcriptomic analysis revealing broad disruption of cholesterol-regulating genes and activation of oxidative stress, ferroptosis, cuproptosis, and apoptosis [27]. Together, these transcriptomic studies illustrate that Prussian blue – based nanoparticles and PTT elicit complex, multi-pathway cellular responses, providing a molecular roadmap for optimizing nanoparticle design and therapeutic synergy.

3.2. Single-cell transcriptomic insights into PBNP-PTT-Induced immune reprogramming

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool for dissecting the cellular and molecular consequences of PTT at single-cell resolution, enabling detailed analysis of PTT-induced shifts within the TME [28]. Using PEGylated iron oxide magnetic nanoparticles – with or without pH-responsive targeting ligands – scRNA-seq has revealed the cellular routes of nanoparticle transport and uptake within solid tumors [29]. These analyses showed that magnetic nanoparticles enter tumors through both vascular bursts and endothelial transcytosis and preferentially accumulate in specific immune subsets, particularly Trem2⁺ regulatory tumor-associated macrophages [29]. ScRNA-seq demonstrated that localized ablative immunotherapy combining PTT with glycated chitosan profoundly reprograms tumor-infiltrating immune cells, increasing interferon-enriched and cytotoxic NK cell states while suppressing inhibitory signaling [30]. In parallel, this treatment expanded multiple γδT cell subtypes, including activated, cytotoxic, interferon-enriched, and IL-17–producing populations [31], supporting enhanced innate and unconventional T cell – mediated antitumor immunity. Although not previously studied for PBNP-PTT, applying single-cell transcriptomics to PBNP-PTT could pinpoint which immune subsets are activated, suppressed, or recruited post-treatment, guiding rational combination strategies with checkpoint inhibitors or cytokine therapies.

3.3. Proteomic profiling of PBNP-PTT-Mediated tumor and immune modulation

Proteomic profiling provides critical insights into the molecular consequences of PBNP-PTT, revealing how thermal stress reshapes cellular signaling, structural integrity, and immune activity. Wang et al., reported a mitochondria-targeting PBNP platform (MiBaMc) composed of biologically precipitated PBNPs integrated with bacterial membrane fragments and mitochondrial-targeting ligands [32]. Proteomic analyses indicate that MiBaMc-mediated phototherapy induces profound mitochondrial protein damage, disrupts oxidative phosphorylation, and activates stress pathways, leading to ICD [32]. These proteomic changes promote tumor antigen release, dendritic cell maturation, and T cell – mediated immunity, resulting in strong synergy with anti-PD-L1 therapy in vivo in murine models of breast and colorectal cancer [32]. In cholangiocarcinoma models, proteomic profiling confirmed transcriptomic results by showing that laser-activated gold-decorated iron oxide nanoflowers combined with PD-1 blockade restored tumor-suppressive proteins, boosted cytokine and interferon signaling, and reduced pro-tumoral pathways [33]. These changes aligned with increased CD8⁺ T-cell and NK-cell infiltration, demonstrating a clear shift toward an immunostimulatory TME [33]. Future proteomic work should pinpoint proteins and pathways that drive thermal response, immune activation, and overall PBNP-PTT efficacy. Mass spectrometry can reveal biomarkers for treatment response and patient stratification. Applying these approaches to PBNP-PTT will clarify how PBNPs modulate protein networks, cytokines, and antigen processing, helping optimize and advance PBNP-based immuno-photothermal therapies.

3.4. Integrative metabolomic and lipidomic insights into PBNP-PTT

Metabolomic studies indicate that PTT can markedly disrupt tumor metabolic homeostasis, creating opportunities for coupling metabolic regulation with antitumor immunity. PTT reshapes key metabolic programs by influencing hypoxia, glycolytic activity, and oxidative stress, while complementary proteomic analyses show alterations in pathways such as pyruvate and choline metabolism that are closely linked to apoptosis induction in cancer cells [34]. A cell membrane – camouflaged PBNP nanoplatform (MPB-3BP@CM NPs) was developed to couple photothermal therapy with metabolic and immune modulation in colorectal cancer [35]. Metabolomic analyses revealed that 3-bromopyruvate delivery inhibited hexokinase II – driven glycolysis, reducing ATP and lactate levels and reprogramming tumor-associated macrophages toward a pro-inflammatory M1 phenotype [35]. When combined with PBNP-mediated PTT, this coordinated metabolic disruption promoted immune activation and significantly enhanced antitumor efficacy, underscoring the value of metabolomics and lipidomics in optimizing immuno-photothermal PBNP strategies [35]. Metabolomic and lipidomic profiling showed that the paclitaxel – molybdenum boride nanosheet platform induces broad metabolic reprogramming in lung cancer, disrupting amino acid balance and lipid synthesis [36]. These changes suggest that improved PTT efficacy arises from both direct cytotoxicity and metabolic network remodeling that helps overcome chemoresistance [36]. Together, these metabolomic and lipidomic insights underscore the value of extending such analyses to PBNP-PTT, where similar metabolic reprogramming could guide optimization of therapeutic efficacy.

Collectively, these multiomics tools reveal how PTT reshapes gene expression, protein networks, and metabolic pathways, and their integration with PBNP research, a current gap in the field, will be essential for decoding PBNP-driven biological responses and refining next-generation photothermal therapies.

3.5. ML-Multiomics as a synergistic closed-loop framework for optimizing PBNP-PTT

Multiomics data are integrated by combining genomics, transcriptomics, proteomics, metabolomics, and related datasets to capture interactions across biological layers. This integration is increasingly driven by ML and AI approaches, including deep learning, graph neural networks, and generative models, that harmonize heterogeneous data, extract shared features, and model complex regulatory relationships (Figure 1) [37]. For a deeper and more comprehensive discussion of AI- and ML – enabled multiomics integration, including methodological advances and remaining challenges, readers are referred to recent dedicated reviews on this topic [37,38].

Applied to PBNP-PTT, a ML-multiomics integration loop could consist of iteratively generating therapeutic perturbations (e.g., PBNP design and laser irradiation regime), profiling multiomic responses, learning predictive structures from the data, and redesigning treatment parameters based on model outputs and predictions. PBNPs are well suited for ML – multiomics integration because their reproducible synthesis and widely described photothermal properties enable the generation of consistent, high-quality multiomic datasets. Consequently, the systematic variation of PBNP design and PTT administration parameters could be readily correlated with transcriptomic, proteomic, metabolomic, and immune responses using ML models yielding a tractable platform for ML-guided, closed-loop optimization of PBNP-PTT. For example, unsupervised ML models could be used to reduce high dimensional multiomics datasets into underlying, coordinated molecular response programs associated with PBNP-PTT while supervised ML models could be trained on known outcomes (e.g., treatment response or immune activation) to predict therapeutic efficacy as a function of various treatment parameters [37,39]. The convergence of ML and multiomics represents a critical next step for PBNP-PTT.

4. Clinical translation considerations for PBNP-PTT

The clinical translational pathway for nanoparticle-mediated PTT has been pioneered primarily through gold nanoparticle-mediated PTT platforms. Gold nanoshells, commercialized as AuroShell® nanoparticles (AuroLase® therapy, Nanospectra Biosciences), have demonstrated that nanoparticle-mediated PTT can progress from preclinical development through regulatory approval to clinical evaluation, possibly informing the future clinical translation of PBNP-PTT.

AuroShell® nanoshells underwent comprehensive Good Laboratory Practice (GLP) compliant preclinical toxicity testing conforming to ISO 10993 standards [40]. Gad et al. conducted GLP studies with intravenous delivery in mice, rats, and beagle dogs over follow-up periods extending to 404 days [40]. No toxicity or biological incompatibility was identified. This rigorous approach established a translational blueprint for comprehensive biocompatibility testing with multi-species studies that could be extended to PBNP-PTT in the future. Stern et al. conducted the first human safety study of nanoparticle-mediated PTT in prostate cancer, demonstrating that AuroShell®-mediated PTT was well tolerated, with no serious adverse events and localized thermal effects confined to the target tissue [41]. Rastinehad et al. conducted the first clinical pilot trial of AuroShell® PTT in men with localized prostate cancer, showing that most treated lesions were tumor-free at 12 months and that prostate function were preserved [42]. Although long-term efficacy was not assessed, the study demonstrated that nanoparticle-directed PTT can safely achieve focal tumor ablation. Early-phase evaluations of gold nanoshells in other cancers further highlight the modality’s versatility [43], suggesting that PBNPs may likewise be adaptable across multiple tumor types.

Despite clinical progress with gold nanoparticle-mediated PTT, PBNP-PTT has not yet entered clinical testing for cancer therapy. Prussian blue does have an established regulatory precedent through FDA approval of oral Radiogardase® to treat radioactive cesium-137 or thallium poisoning [44]. However, it is critical to emphasize that oral Prussian blue and intravenously (i.v.) or intratumorally administered PBNPs are pharmacokinetically and toxicologically distinct. While the established safety record of oral Prussian blue provides a starting point for clinical translation for PBNP-PTT, this platform still requires full GLP compliant toxicology studies including dose escalation, long-term biodistribution, organ retention, immunogenicity, and clearance analyses prior to clinical translation.

4.1. Critical quality attributes of PBNPs

For nanoparticle platforms including PBNPs, successful clinical translation depends on the identification, characterization, and control of critical quality attributes (CQAs) that influence their safety, efficacy, manufacturability, and regulatory approval (listed in Table 2). From a translational perspective, gold nanoparticles benefit from chemical inertness, negligible biodegradation and extensive historical toxicology datasets, all of which facilitated their advancement into first-in-human PTT trials [45]. However, this inertness can lead to long term tissue retention, particularly in the liver and spleen, raising concerns regarding chronic accumulation and delayed clearance [46]. In contrast, PBNPs exhibit biodegradability and pH-dependent dissolution [47], which may enable metabolic clearance but also introduces additional safety considerations related to transient organ burden, oxidative stress, inflammatory responses and dose dependent toxicity [48]. Therefore, PBNPs offer potential advantages in biodegradability and clearance but require rigorous exploration of degradation behavior, dose dependent toxicity, and immunological responses to achieve comparable clinical readiness to gold nanoparticles.

Table 2.

Critical quality attributes (CQAs) of Prussian blue nanoparticles and their functional and translational implications.

Critical Quality Attribute (CQA)	Physicochemical or Biological Role	Impact on PTT Performance	Translational/Manufacturing Significance
Particle size & PDI	Governs colloidal behavior and vascular transport	Tumor penetration, heat distribution uniformity	Batch reproducibility, stability during storage
Zeta potential	Surface charge and protein corona formation	Circulation time, immune recognition	Formulation robustness, aggregation control
Crystallinity & defect density	Determines electronic structure and phonon behavior	Photothermal conversion efficiency, thermal stability	Lot-to-lot consistency in heating performance
Optical absorption peak characteristics	Defines NIR light absorption efficiency	Depth of heating and laser – particle matching	Device – drug compatibility (laser wavelength selection)
Photothermal conversion efficiency	Heat generation per unit optical input	Therapeutic dose efficiency and safety window	Dosing standardization across patients
Surface chemistry	Biological interface with cells and proteins	Biodistribution, uptake, immune interaction	Scalable functionalization strategies
Degradation kinetics	Structural stability in physiological environments	Duration of heating capability	Clearance profile, long-term safety
Endotoxin contamination	Immunostimulatory impurity	Inflammatory confounding effects	GMP release criteria
In vivo biodistribution & clearance	Whole-body fate and organ accumulation	Off-target heating risk	Regulatory safety assessment, dosing limits

In terms of surface chemistry both PBNPs and gold nanoparticles require surface functionalization to achieve colloidal stability, control zeta potential and modulate pharmacokinetics [49,50]. PEGylation strategies have been widely adopted for both platforms [49,50]. However, PBNPs additionally face challenges related to surface mediated dissolution, and endotoxin adsorption, necessitating coatings that simultaneously stabilize the crystal lattice, preserve optical absorption and minimize immunogenic contamination [50,51]. Overall, gold platforms offer exceptional stability and established clinical precedent but have long term persistence in the body, whereas PBNPs provide biodegradability and other benefits such as intrinsic iron-based imaging compatibility [52]. PBNPs require further physicochemical and degradation related CQAs to bolster clinical translation.

4.2. Nanoformulation considerations for PBNP-PTT

A foundational requirement for clinical translation of PBNP-PTT is the establishment of scalable, reproducible synthesis protocols for the nanoparticles conforming to good manufacturing practice (GMP) standards, consistent with standardized physicochemical and lot-to-lot characterization principles emphasized by the agencies such as the National Cancer Institute (NCI)’ s Nanotechnology Characterization Laboratory (NCL) Assay Cascade Protocols [53]. GMP manufacturing demands rigorous consistency, validated analytical methods and comprehensive quality control documentation [53]. For PBNPs, this entails precise control over nanoparticle size distribution, crystallinity, morphology, surface charge and surface functionalization, as these parameters directly influence optical absorption, photothermal conversion efficiency, colloidal stability and biological interactions [54]. Even minor variations in reactant concentrations, temperature or pH during synthesis can produce nanoparticles with varied physicochemical properties, potentially compromising therapeutic performance or safety [55].

Although Prussian blue is FDA-approved, the safety profile of orally administered micron-sized Prussian blue particles cannot directly inform the safety of i.v. or intratumorally administered PBNPs [56]. In a study by Qu et al., although i.v. administration of PBNPs at lower doses (5–10 mg/kg) showed no obvious toxicity in mice, higher doses (20 mg/kg) caused transient appetite loss and weight reduction, with significant particle accumulation in the liver and lungs that triggered inflammatory responses and oxidative stress [56]. Consistent with these organ level findings, Chen et al. previously reported that i.v. injected PBNPs induce acute but reversible biological disturbances, including transient elevations in liver injury markers and short-term reductions in T-cell frequencies in the blood and spleen, with recovery to baseline occurring within days to weeks [48]. Furthermore, while studies by Cano-Mejia et al. and Doveri et al. demonstrated that PBNPs undergo pH dependent degradation at physiological pH and that polymeric coatings or protein corona formation can stabilize PBNPs and prevent rapid dissolution, the long-term clearance profiles and comparative metabolism of coated versus uncoated PBNPs via hepatic and renal routes remain incompletely characterized and require systematic investigation [47,57]. Therefore, in alignment with the standardized physicochemical and in vivo characterization principles outlined in the NCI’s NCL Assay Cascade, comprehensive GLP-compliant toxicology studies that cover acute and chronic exposure, biodistribution, degradation and immunotoxicity are essential to establish a preclinical safety foundation comparable to that achieved for gold nanoshells [53].

Immunogenicity and hypersensitivity reactions represent additional safety considerations. While PBNPs are generally considered biocompatible, individual patient factors, including apolipoprotein phenotypes, serum lipid profiles, and baseline immune status, can modulate nanoparticle protein interactions and potentially influence the risk of complement activation-related pseudoallergy, cytokine release, or antibody-mediated clearance [58]. However, recent evidence from Betrand et al. suggests that complement activation alone may not be the sole predictor of nanoparticle circulation time, as both short- and long-circulating PEGylated nanoparticles show similar clearance patterns in complement-deficient mice models [58]. The complex interplay between nanoparticles and innate immunity cascades remains incompletely understood, particularly regarding clearance after repeated dosing and the development of pseudo-allergic reactions in patients [59]. These gaps further underscore the importance of adopting NCL-aligned immunotoxicity assessments, including complement activation assays, cytokine profiling and repeated-dose immunological monitoring, to guide safe nanoformulation design for clinical PBNP-PTT.

4.3. Laser considerations for PBNP-PTT

The laser is an integral part of PTT. Light absorbing nanoparticles localized in the tumor convert optical energy into heat. Laser parameters such as wavelength, power density, pulse structure (continuous wave versus pulsed), and exposure time all influence the heating rate, spatial temperature distribution and the risk of collateral tissue damage associated with the treatment. Failure to control these parameters precisely may result in ineffective ablation or unintended thermal injury to adjacent structures. Therefore, translating PBNP -PTT into the clinic requires full specification of laser parameters.

Tissue penetration of the laser is a major determinant of whether PBNP-mediated PTT can effectively reach the targeted tumor region. Biological tissues exhibit wavelength dependent absorption and scattering due to endogenous chromophores (such as hemoglobin, melanin, water, lipids) as well as to structural heterogeneity such as cell membranes, collagen fibers and other microarchitectural features [60]. As a result, the depth at which sufficient photon flux can be delivered to a tumor, and thus the feasibility of achieving cytotoxic heating (e.g., ≥ 43°C for several minutes), depends critically on the choice of laser wavelength, power density, beam geometry, pulse structure (continuous vs. pulsed), and exposure duration [61]. The NIR optical window, particularly NIR-I (~700–900 nm), minimizes absorption and scattering and is therefore standard in preclinical PTT; 808 nm is widely used due to commercial laser availability and compatibility with nanoparticle absorption profiles [62,63].

For current PBNP formulations, strong charge transfer absorption centered in the NIR-I region (typically ~700–750 nm with extended tails toward ~800 nm) supports the use of clinically available 808 nm diode lasers as a practical and sufficient excitation source in many settings [12]. Several studies have demonstrated efficient photothermal heating of PBNPs under 808 nm continuous wave irradiation at power densities on the order of ~0.3–2 W·cm⁻², achieving therapeutically relevant temperature elevations without requiring specialized laser hardware [9,12,64,65]. Accordingly, existing FDA-cleared 808 nm diode laser platforms used in other thermal therapies could be repurposed for early PBNP-PTT clinical studies, subject to indication specific Investigational Device Exemption (IDE) approval. However, even in NIR-I, residual absorption and scattering constrain therapeutic penetration, typically ≤3 cm depending on tissue composition [62,66]. Furthermore, penetration depth defined as 1/e intensity does not ensure adequate thermal dose at depth, as heterogeneous optical properties, perfusion, and heat sinks decouple surface fluence from tumor temperature rise [67].

For deeper tumors or anatomically challenging sites, absorption tuning of PBNPs, such as through metal substitution, defect engineering, or hybridization to shift or broaden NIR absorption may be required to better match available clinical laser wavelengths or to enable lower power operation [9,12,68]. Beyond wavelength selection, continuous wave irradiation is currently the most studied mode for PBNP-PTT translation. Whether PBNP-PTT can benefit from nanosecond or millisecond pulsed regimes, as demonstrated for nanoparticles like gold, remains an open question and requires dedicated characterization of PBNP thermal relaxation kinetics and pulse width dependent heating efficiency.

Beyond the biophysical and biological challenges, clinical translation of a PBNP-PTT laser system requires regulatory compliance. In the United States, a novel therapeutic laser system constitutes as a medical device, and its use in human subjects requires an IDE under 21 CFR Part 812, along with IRB approval and informed consent. The IDE submission must include a full device description that incorporates details such as laser wavelength, output power, pulse structure, beam geometry plus safety features, calibration procedures, non-clinical data, risk analysis and a clinical investigation plan demonstrating acceptable risk – benefit.

Given the significant risk classification of therapeutic laser devices, FDA Notified Bodies require a complete technical, safety, and quality dossier prior to authorizing clinical use.

4.4. Imaging considerations for theranostic PBNP -PTT

The translation potential of PBNP-PTT extends beyond single modality treatments toward integrated theranostic platforms combining therapy with image guidance. Unlike nanomaterials requiring extensive functionalization or exogenous contrast agents, PBNPs possess intrinsic multimodal imaging capabilities. Their iron-based composition enables magnetic resonance imaging (MRI) while strong NIR supports both PTT and photoacoustic imaging (PAI) [69,70]. Foundational work by Shokouhimehr et al., demonstrated inherent T₁- and T₂-weighted MRI visibility of PBNPs [71]. Building on this, Dumont et al., demonstrated that manganese-containing PBNPs can serve as molecular MRI and fluorescence imaging agents for pediatric brain tumors, enabling in vitro imaging and ex vivo biodistribution analysis in an orthotropic mouse model [72]. These studies established the feasibility of multimodal PBNP imaging and integration into MRI-guided therapeutic workflows prior to PTT, although real-time MRI guidance during PTT may be constrained by high cost and logistical limitations.

PAI offers high spatial resolution and penetration depth while enabling realtime feedback during PTT by converting absorbed optical energy into acoustic signals [73]. Cheng et al. demonstrated that PEGylated PB nanocubes provided strong contrast in both T₁ weighted MRI and photoacoustic tomography, enabling multimodal imaging-guided tumor localization and treatment monitoring [69]. Similarly, Xu et al. reported concentration-dependent PAI signals from optimized PBNPs, with marked in vivo tumor signal enhancement following intravenous administration [70]. Combining MRI for anatomical localization and treatment planning with PAI for real time monitoring enables a comprehensive theranostic workflow; improved PBNP dispersibility enhanced T₁ MRI and PAI signals, enabling visualization of nanoparticle accumulation prior to laser activation and subsequent PTT achieved extensive tumor necrosis and growth suppression [70]. Despite these promising results, PAI-guided PTT remains largely preclinical and broader translation will depend on device availability, regulatory IDE clearance, and integration into clinical workflows.

More cost-effective imaging modalities like ultrasound imaging may offer a scalable alternative for image guided PBNP-PTT, especially for interstitial nanoparticle delivery. Ultrasound‐guided interstitial PTT with PBNPs improves nanoparticle injection precision and laser-fiber placement, leading to enhanced tumor regression and long-term survival in a mouse neuroblastoma model [74]. Although PBNPs are not directly visible under ultrasound, visualization of needles, catheters and the surrounding tissue anatomy enables accurate targeting. Emerging hybrid ultrasound-PAI platforms further extend this approach by combining real-time guidance with enhanced catheter tracking, thermometry, and lesion monitoring [75]

Clinical translation of PBNP-PTT will require the same rigorous framework established for gold-nanoshell PTT, including GMP-standard nanoformulation, GLP toxicology, and well-defined laser device specifications. Although Prussian blue has an FDA-approved safety foundation, PBNPs still demand comprehensive evaluation of biodistribution, degradation, immunotoxicity, and laser-tissue interactions. Finally, multimodal imaging such as MRI, PAI, and ultrasound guidance can be integrated to support accurate nanoparticle delivery and treatment monitoring, strengthening the translational pathway for PBNP-PTT.

5. Forward-looking applications of PBNP-PTT

Looking ahead, several emerging approaches – including laser optimization, advanced PBNP designs, multiomics – driven insights, and synergistic combination strategies – are poised to significantly enhance the capabilities of PBNP-PTT (Figure 2).

Figure 2.

Forward-looking strategies to enhance PBNP-PTT. Schematic representation of emerging innovations designed to advance the performance and translational potential of PBNP-PTT. These include: (i) laser optimization using NIR-II irradiation to improve tissue penetration and thermal precision; (ii) nanoparticle engineering strategies such as biomimetic PBNPs with cell membrane coatings; (iii) analytical profiling tools including single-cell TCR sequencing (scTCR-seq) for monitoring T cell clonality and immune activation and immunopeptidomics for mapping PTT-induced antigen presentation; (iv) combination therapy modalities incorporating extracellular matrix (ECM) modulation and CAR T cell therapy. Collectively, these forward-looking strategies highlight the potential for PBNP-PTT to evolve into an integrated, immune-driven cancer nanomedicine. Created in BioRender. Sweeney, L. (2026) https://BioRender.com/qs0xast.

5.1. Next-generation laser technologies for precision PTT

Traditionally, photothermal agents operating in the first near-infrared window (NIR-I) have been widely studied due to their favorable light absorption and moderate tissue penetration [1]. However, as described earlier, their clinical translation is limited by strong light scattering and shallow penetration depth in biological tissues. In contrast, NIR-II (~900–2000 nm) offers superior optical advantages, including deeper tissue penetration, reduced photon scattering, higher permissible exposure limits, and enhanced photothermal conversion efficiency making it particularly suitable for treating deep-seated solid tumors [76]. To fully exploit these benefits, PBNPs must be engineered to exhibit strong and tunable absorption within the NIR-II region. This can be achieved through strategies such as doping with transition metals (e.g., Cu, Mn, Co), hybridizing with NIR-II absorbing materials (e.g., gold nanorods or rare-earth dopants), or modulating crystal structure and defect density to extend absorption spectra. Prussian blue@polyacrylic acid/copper sulfide Janus nanoparticles (PB@PAA/CuS JNPs) were designed to enhance photothermal conversion efficiency by engineering energy-level transitions within a Janus structure [77]. This architecture enables strong absorption in the NIR-II window, leading to deeper tissue penetration under 1064 nm laser irradiation compared with NIR-I excitation [77]. As a result, PB@PAA/CuS JNPs achieve improved photothermal tumor inhibition and show particular promise for treating deep-seated tumors, highlighting NIR-II – optimized nanostructural design as a key strategy for advancing PTT [77]. A PB – based nanoplatform (PB@Ag₂S) was developed for glioma therapy, leveraging NIR irradiation to enhance CDT, transiently disrupt the blood – brain barrier (BBB), and increase intracranial nanoparticle accumulation [78]. The Ag₂S component enables NIR-II fluorescence imaging for real-time monitoring [78]. Beyond tumor ablation, the PBNP system reprograms the immunosuppressive glioma microenvironment by promoting dendritic cell maturation and antitumor macrophage polarization, leading to effective tumor eradication [78]. This work highlights the potential of NIR-II – enabled PBNP theranostics for treating deep-seated brain tumors. Developing PBNPs optimized for NIR-II excitation would significantly enhance their therapeutic precision, imaging capabilities, and translational potential for next-generation photothermal nanomedicine.

5.2. Optimizing nanoparticle composition for enhanced PTT performance

Efficient tumor targeting remains difficult for i.v. delivered nanoparticles due to opsonization, macrophage clearance, and poor vascular extravasation. Beyond PEGylation and ligand targeting, biologically inspired “disguised” nanocarriers such as albumin-based systems, extracellular vesicles (EV), and cell membrane – coated nanoparticles offer more advanced solutions [79]. Albumin nanoparticles provide biocompatibility, long circulation, and strong tumor accumulation, enabling multimodal imaging and chemo-photothermal therapy [80,81]. Natural and engineered EVs, including those from NK [82], dendritic cells [83], and macrophage [84], deliver inherent cytotoxic and antigen-presenting functions that enhance photo-immunotherapy. Biomimetic PBNP coated with MnO₂ and camouflaged by RGD-modified erythrocyte membranes were designed to overcome tumor hypoxia and redox barriers [85]. This coating enabled oxygen supply, glutathione depletion, enhanced tumor targeting, and immune evasion, resulting in improved PDT/PTT synergy and strong tumor inhibition in vivo, highlighting biomimetic PBNPs as effective TME-responsive theranostic platforms [85]. A glioblastoma cell – membrane – coated PBNP (Exo:PB) enables BBB penetration, tumor-specific accumulation, PAI, and effective NIR photothermal ablation [65]. This biomimetic platform outperforms conventional PBNPs, demonstrating targeted, noninvasive theranostic potential for glioblastoma [65]. Cell membrane – coated nanoparticles inherit immune-evasive, tumor-homing traits [86], while virus-like particles offer safe capsid structures for drug delivery and immune activation [87]. Collectively, these biomimetic nanocarriers represent a rapidly advancing strategy to improve tumor targeting, sharpen PBNP-PTT precision, and strengthen synergistic therapeutic responses.

Multi-stimuli – responsive nanoparticles are advancing PBNP-PTT by enabling activation that is tightly confined to the TME [88]. These adaptive systems exploit endogenous cues such as acidity, hypoxia, redox imbalance, and ROS or external triggers like light and heat to overcome key delivery and selectivity barriers [88]. A dual-responsive PB – based nanoplatform was developed for targeted theranostic treatment of triple-negative breast cancer [89]. The system enables pH- and glutathione-triggered drug release, enhances tumor targeting via hyaluronic acid, and delivers synergistic chemo-photothermal therapy that induces TP53-related apoptosis [89]. Doped PBNPs also provide strong T₁/T₂ MRI contrast, supporting image-guided and personalized cancer therapy [89]. Together, these strategies show how rationally engineered, stimulus-responsive nanomaterials can markedly enhance the precision and therapeutic performance of PTT – and provide a strong design framework for developing next-generation, PBNP-based photothermal platforms.

Biodegradable PB analogues show strong antitumor efficacy through synergistic PTT/PDT/CDT while maintaining low in vivo toxicity [90]. Their intrinsic biodegradability and epithelial – mesenchymal transition modulation confers anti-metastatic effects, and built-in MRI/PA imaging supports their use as safe, multifunctional cancer theranostic platforms [90]. Biodegradable nanoparticles are an important direction for future PBNP derivatives, as degradable coatings and hybrid designs improve systemic clearance and reduce long-term accumulation – key considerations for regulatory approval [91]. Existing biodegradation-friendly platforms span synthetic polymers (PLGA, PLA, PCL, PEG), natural polymers (chitosan, alginate, gelatin, dextran, hyaluronic acid), and protein carriers like HSA and BSA [91]. Lipid-based systems such as liposomes, SLNs, and NLCs offer intrinsic biocompatibility, while inorganic options – including degradable MSNs, biocompatible MOFs, and calcium-based nanoparticles – further broaden available chemistries [91]. Collectively, these materials provide safe, versatile, and clinically relevant foundations for controlled therapeutic and diagnostic delivery.

5.3. Decoding tumor biology with omics to direct precision PTT

Single-cell TCR sequencing (scTCR-seq) can offer a high-resolution view of how PBNP-PTT reshapes adaptive immunity, revealing T-cell clonal expansion patterns that bulk omics cannot capture [92]. By mapping TCR clonotypes, scTCR-seq can identify the expansion of tumor-reactive CD8⁺ T-cell clones after PBNP-PTT, potentially serving as a biomarker of effective antigen release and immune activation. When paired with scRNA-seq, it links clonal dynamics to functional states – such as cytotoxicity, memory formation, or exhaustion – providing insight into how therapy transforms the TME [92]. Applying these tools to PBNP-PTT may uncover whether treatment recruits new tumor-reactive clones and enhances local immune activity, guiding the rational design of future PTT-immunotherapy combinations. Immunopeptidomics provides a powerful way to define how PBNP-PTT alters the tumor antigen landscape [93] by identifying new MHC-I – presented peptides generated through thermal stress and immunogenic cell death [94].

Recent preliminary studies reported that PBNP-PTT of a GBM cell line reshapes tumor antigen release and presentation, with immunopeptidomics identifying a distinct MHC-I peptide repertoire in PTT-primed dendritic cells and scTCR-seq revealing increased clonality and selective expansion of multiple tumor-specific CD8⁺ TCR clones (ASGCT 2025 Abstract #1227) [95]. Applying similar analyses to PBNPs could reveal whether PTT expands the repertoire of immunogenic epitopes and strengthens systemic immune activation. These insights could guide the development of peptide vaccines, nanovaccines, and PTT – checkpoint inhibitor combinations, and, when paired with scTCR-seq, map neoantigen-driven T-cell expansions, positioning PBNP-PTT as a potential in situ vaccine for durable anti-tumor immunity.

5.4. Future directions for optimizing PTT-Based combination therapies

PTT can remodel the TME by disrupting dense extracellular matrix (ECM) stromal barriers, as hyperthermia degrades collagen and improves drug and immune cell penetration [74]. In pancreatic ductal adenocarcinoma, acid-responsive C-G nanoparticles co-delivering photothermal agents and gemcitabine reduce fibrosis by suppressing transforming growth factor β (TGF- β) signaling in pancreatic stellate cells, enhancing extracellular matrix breakdown and boosting chemophotothermal efficacy [96]. PBNP remodel the immunosuppressive oral squamous cell carcinoma TME by reducing TGF-β signaling and reprogramming tumor-associated macrophages from pro-tumor M2 to anti-tumor M1 phenotypes [24]. Similar benefits are seen in other desmoplastic tumors: combining PTT using gold-decorated iron oxide nanoflowers with PD-1 blockade in cholangiocarcinoma decreases stromal stiffness, reprograms fibroblasts, and increases CD8⁺ T-cell infiltration, converting immune-cold tumors into immune-responsive ones [20]. Collectively, these findings highlight PBNP-PTT’s potential to modulate stromal architecture and improve the effectiveness of both cytotoxic and immunotherapeutic strategies in highly fibrotic cancers.

Multimodal PBNPs represent a sophisticated evolution of the platform, where the nanoparticle is engineered to perform diagnostic imaging, targeted therapy, and microenvironment regulation simultaneously. By integrating additional elements (like Manganese Mn²⁺ or Gadolinium Gd³⁺) or coating them with bioactive membranes, these particles move beyond simple photothermal agents to become “all-in-one” theranostic tools to overcome the limitations of monotherapies, improving efficacy, reducing toxicity, and helping bypass resistance mechanisms [97]. Standard PBNPs are already effective for PAI because they absorb NIR light and convert it into ultrasonic waves. However, multimodal PBNPs are doped or modified to allow for deep-tissue visualization. By doping the PBNP lattice with paramagnetic ions like Gd³⁺ or Mn²⁺, these nanoparticles provide high-contrast MR imaging [98]. Doping with high-Z elements (like Gold or Bismuth) allows the PBNPs to act as contrast agents for X-ray CT. If doped with Mn or Copper (Cu), the nanoparticles can trigger a Fenton-like reaction [99]. They convert the tumor’s endogenous hydrogen peroxide (H₂O₂) into highly toxic hydroxyl radicals (OH), killing cancer cells via oxidative stress [99]. As co-delivery technologies continue to advance, multimodal nanoparticle systems hold strong promises for enabling integrated combination therapies, though challenges in large-scale manufacturing, pharmacokinetics, and regulatory approval remain important considerations.

PTT is increasingly integrated with advanced immunotherapies – such as CAR-T cells, oncolytic viruses, and cancer vaccines – to overcome resistance in solid tumors and strengthen antitumor immunity. PBNP-PTT functions as an in-situ vaccine by inducing ICD and releasing tumor antigens [2,100], creating a microenvironment that supports adoptive and viral therapies. Mild hyperthermia can remodel dense stroma, improve perfusion, and enhance immune infiltration [1,101], thereby boosting CAR-T cell activity, as seen with improved efficacy of CSPG4-CAR-T cells after photothermal ablation [102]. Similarly, combining PTT with oncolytic viruses enhances viral delivery, spread, and immune activation, as demonstrated by platelet-membrane – coated, ICG-loaded vaccinia virus platforms [103]. PTT also strengthens cancer vaccination, where targeted photothermal activation improves antigen presentation, systemic immunity, and memory responses, especially when paired with checkpoint blockade [104].

Collectively, these forward-looking applications illustrate the potential of PBNPs to evolve from local tumor ablation agents into multifunctional tools for precision cancer immunotherapy. By integrating advances in nanotechnology, imaging, immunology, and computational modeling, PBNPs are poised to play a transformative role in next-generation cancer treatments, offering both improved efficacy and personalized therapeutic strategies (Figure 2).

6. Conclusion

PBNPs are emerging as one of the most versatile and clinically adaptable platforms for PTT, uniting established biocompatibility with expanding opportunities in cancer immunotherapy. As the field advances, several converging domains presented here collectively define the roadmap for next-generation of PBNP-PTT mono and combination therapies.

ML approaches are poised to reshape how PBNP-PTT is designed, optimized, and personalized. Multimodal ML models can predict optimal nanoparticle properties, laser parameters, and treatment regimens based on imaging, physicochemical, and clinical input data. Mechanistically informed ML approaches will help clarify the key nanoparticle and tumor features that govern therapeutic efficacy and immune activation, while large-scale collaborative data integration will support the development of more robust and generalizable predictive models. Ultimately, advanced computational simulations powered by ML, capable of modeling PBNP biodistribution, thermal behavior, and immune responses, may enable virtual treatment testing and guide individualized therapy planning, accelerating translation toward precision oncology.

Complementing computational advances, multiomics provides a systems-level understanding of how PBNP-PTT reshapes tumor biology and immunity. By integrating transcriptomics, proteomics, metabolomics, and single-cell sequencing, researchers can elucidate the molecular programs activated by photothermal stress, identify biomarkers of response, and uncover rational combination strategies with immunotherapy. Immunopeptidomics extends this framework by mapping PTT-induced changes in tumor antigen presentation, supporting the development of in situ vaccination strategies or personalized cancer vaccines, while single-cell TCR sequencing enables tracking of clonal T-cell dynamics to guide pairing with checkpoint blockade or adoptive cell therapies. Emerging analytical technologies – including spatial immune mapping, high-throughput antigen discovery, and advanced immune profiling – provide expanded tools to monitor PTT responses and better predict therapeutic outcomes. Together, these integrated analytical approaches elevate PBNP-PTT from a primarily cytotoxic modality to a precision immune-modulating platform capable of informing individualized cancer therapy.

Clinically, nanoparticle-mediated PTT has entered a new phase, with gold nanoshells demonstrating the feasibility and safety of thermally activated nanomaterials in humans. PBNPs stand to benefit from these precedents, bolstered by their FDA-recognized safety profile and scalable manufacturing. Key challenges – including controlling biodistribution, managing long-term retention, and accounting for interpatient variability in immune responses – must be systematically addressed through image-guided delivery, standardized laser protocols, and biomarker-driven trial designs. With these strategies, PBNPs are well positioned to enter early-phase clinical evaluation as a next-generation immuno-photothermal therapeutic.

7. Future perspective

Looking forward, the future of PBNP-PTT extends beyond tumor ablation toward integrated, multifunctional, and adaptive cancer therapy. Emerging directions include biodegradable or stimuli-responsive PBNP derivatives that enhance clearance and specificity; “disguised” nanoparticles cloaked with cell membranes or immune-evasive coatings to improve tumor homing and reduce off-target clearance; theranostic formulations that unite imaging, drug delivery, and immunomodulation; and synergistic combinations with CAR-T cells, oncolytic viruses, vaccines, and immune checkpoint inhibitors. These innovations promise to amplify both local tumor destruction and systemic antitumor immunity, positioning PBNPs as a central component in next-generation immunotherapy strategies. In summary, PBNPs offer a uniquely promising foundation for the evolution of PTT from empirical heating to precision immune engineering. By integrating AI/ML, systems-level biological profiling, clinical translation strategies, and forward-looking nanoparticle design, future PBNP-based systems can deliver safer, more effective, and more personalized cancer treatments, ultimately helping to shape the next generation of nanomedicine and cancer immunotherapy.

Funding Statement

This manuscript was funded by the Fischell Department of Bioengineering at the University of Maryland College Park, the Edward and Jennifer St. John Center for Translational Engineering and Medicine, and the National Cancer Institute of the National Institutes of Health under grant R01CA290045. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Article highlights

• Prussian blue nanoparticles (PBNPs) represent a promising translational photothermal therapy (PTT) platform for cancer and PBNP-PTT has demonstrated capacity to induce immunogenic cell death and immune activation beyond local tumor ablation under near-infrared (NIR) laser irradiation.

• Machine learning (ML) provides a conceptual framework to optimize PBNP-PTT across different variables such as nanoparticle design, dosing, laser parameters and irradiation regime but PBNP specific training data sets are currently limited.

• ML strategies developed for other nanoparticles such as gold offer transferable precedent to PBNP-PTT optimization but require adaptation to capture PBNP-specific relationships between structure and function.

• Integrated multiomics profiling (transcriptomics, proteomics, and metabolomics) of solid tumors after PBNP-PTT demonstrates concerted molecular reprogramming, marked by enhanced pathways linked to immunogenic cell death, antigen processing and presentation, dendritic cell maturation, and T-cell activation, together with altered tumor metabolic homeostasis and suppression of tumor growth and immunosuppressive signaling.

• ML and AI approaches enable the integration of genomics, transcriptomics, proteomics, and metabolomics to capture cross-layer biological interactions. Applied to PBNP-PTT, ML – multiomics can support a closed-loop framework in which nanoparticle design and laser parameters are iteratively linked to molecular and immune responses.

• Clinical translation of PBNP-PTT can build on gold nanoparticle PTT precedent while still requiring PBNP specific evaluation of critical quality attributes, toxicity and clearance profiles.

• Theranostic integration of PBNP-PTT with MRI, photoacoustic imaging and ultrasound shows strong potential but requires regulatory clearance before adoption into clinical workflow.

• Next-generation PBNP-PTT combines NIR-II photothermal agents, biomimetic designs, and multimodal imaging to remodel the tumor microenvironment, including reducing TGF-β, reprogramming TAMs, degrading stromal barriers, and enhancing immune cell infiltration thus providing a strong rationale for combination with immunotherapies.

• Analytical profiling tools such as immunopeptidomics to map PTT-induced antigen presentation and single-cell TCR sequencing (scTCR-seq) to monitor T-cell clonality and immune activation have the potential to inform the design of PBNPs for precise immunotherapy.

Author contributions

Conceptualization – VFM, NS, RF, Funding acquisition – RF, Investigation – VFM, NS, RF, Resources – RF, Supervision – RF, Writing – original draft – VFM, NS, RF, Writing – review & editing – VFM, NS, RF.

Disclosure statement

Rohan Fernandes (RF) is a co-founder of ImmunoBlue, a biotechnology company focused on developing PBNP-based nanoimmunotherapies. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.