Trends and Patterns in Cancer Nanotechnology Research: A Survey of NCI’s caNanoLab and Nanotechnology Characterization Laboratory

Abstract

Cancer nanotechnologies possess immense potential as therapeutic and diagnostic treatment modalities and have undergone significant and rapid advancement in recent years. With this emergence, the complexities of data standards in the field are on the rise. Data sharing and reanalysis is essential to more fully utilize this complex, interdisciplinary information to answer research questions, promote the technologies, optimize use of funding, and maximize the return on scientific investments. In order to support this, various data-sharing portals and repositories have been developed which not only provide searchable nanomaterial characterization data, but also provide access to standardized protocols for synthesis and characterization of nanomaterials as well as cutting-edge publications. The National Cancer Institute’s (NCI) caNanoLab is a dedicated repository for all aspects pertaining to cancer-related nanotechnology data. The searchable database provides a unique opportunity for data mining and the use of artificial intelligence and machine learning, which aims to be an essential arm of future research studies, potentially speeding the design and optimization of next-generation therapies. It also provides an opportunity to track the latest trends and patterns in nanomedicine research. This manuscript provides the first look at such trends extracted from caNanoLab and compares these to similar metrics from the NCI’s Nanotechnology Characterization Laboratory, a laboratory providing preclinical characterization of cancer nanotechnologies to researchers around the globe. Together, these analyses provide insight into the emerging interests of the research community and rise of promising nanoparticle technologies.

Graphical Abstract

1. Introduction

1.1. Cancer nanotechnology landscape

With the remarkable decline in cardiovascular disease-associated mortality rates over the past century, cancer has arisen as the leading cause of noncommunicable disease-related premature death in many countries and ranks among the top three in more than 175 countries [1]. In 2020 alone, 19.3 million new cancer cases and nearly 10.0 million cancer deaths were estimated worldwide [2]. Cancer treatment has progressed from traditional surgical resection, chemotherapy, and radiation approaches to newer modalities such as hormone therapy, immunotherapy, and targeted therapy. Unfortunately, high systemic toxicities, poor pharmacokinetics, low drug delivery efficiencies, and a lack of early diagnostic methods continue to hamper overall treatment effectiveness.

Over the past few decades, significant developments in nanotechnology and its use in medicine have taken place. Well-designed nanoparticles have shown great potential as nanomedicines that can address the current challenges associated with cancer prevention, diagnosis, and therapy. While no universal definition of nanoparticles exist, the definitions of the National Nanotechnology Initiative (i.e., size in at least one dimention between 1 and 100 nm) and the US Food and Drug Administration (FDA) (i.e., size in at least one dimention less than 1000 nm and related to unique size-attributed properties) are commonly recognized [3]. The unique characteristics of nanoparticles, e.g., high surface-to-volume ratio, enhanced carrying capacity, and the ability to prolong systemic circulation while affording protection to the therapeutic cargo, have expanded the variety of therapeutics that may be delivered to a target site and improved overall pharmacokinetics of systemically administered treatments. Furthermore, nanomaterials’ tunable optical, electronic, and magnetic properties enhance the signal intensity and contrast necessary for bioimaging [4]. Since the first nanoparticle study undertaken by Richard Adolf Zsigmondy in 1909 [5], more complicated nanoparticle formulations have been developed to address the complex clinical challenges associated with treatment of cancers. Integrated multifunctional systems with therapeutic agents (including small molecules, peptides and nucleic acids), targeting molecules, imaging modalities, and inert polymers to extend systemic circulation times are all being developed as cancer nanomedicines [6–12].

Among the numerous categories of nanoparticle platforms being investigated for cancer treatment, liposomes have displayed outstanding performance and comprise the majority of FDA-approved nanomedicines for chemotherapy, including Doxil^® (and several generic versions), Marqibo^®, Daunoxome^®, Onivyde^®, and Vyxeos^® [13]. Nanoparticle-mediated diagnostic imaging for cancer has been slower to expand. Although technetium colloids and off-label use of iron oxides have been routinely used for decades, the field has seen little advancement due in part to more stringent regulatory hurdles, lack of tissue/organ selectivity and targeting, manufacturing challenges, and a generally reduced interest from pharma stemming from lower profit margins, among others [14]. Most recently, a carboxydextran-coated iron oxide nanoparticle (Magtrace; Endomagnetics, Ltd.) was approved by the FDA in 2018 to guide sentinel lymph node biopsies for breast cancer patients, eliminating the need for radioactive tracers [15].

Cancer is a systemic disorder originating from altered function of multiple cellular and molecular mechanisms in more than one organ or tissue; one of the systems malfunctioning in cancer is the immune system. In health, the immune system patrols the body to eliminate transformed cells. Tumors form when the immune system fails to provide an effective cancer surveillance. Therefore, significant attention in current cancer therapy is paid to restoring and improving the immune system’s ability to effectively recognize and eliminate transformed cells. Since its original introduction in the late 1800s, immunotherapy of cancer has seen ups and downs with limited efficacy, tumor resistance, and side effects being among the most significant hurdles. The recent advances in nanotechnology have helped to overcome many of these challenges. Major barriers to effective cancer immunotherapy and approaches to overcoming them using nanotechnology are extensively discussed elsewhere [16–20].

Nanoparticles are also being explored for a variety of other therapy-related treatment strategies. For example, the magnetic iron oxide nanoparticle NanoTherm (MagForce USA) is currently in clinical trials in the U.S. for use in conjunction with an alternating magnetic field to heat and kill prostate cancer cells (NCT05010759). Hensify is approved in the EU for locally-advanced soft tissue sarcoma and is in clinical trials in the U.S. (NCT04505267, NCT04484909, NCT04862455, NCT05039632, NCT04834349, NCT04615013). This hafnium oxide nanoparticle acts as a radioenhancer to amplify the killing effect of radiation on tumor cells. Other areas for which cancer nanotechnology appears highly promising include immunotherapy, gene therapy, cancer vaccines and personalized medicine. For instance, a lipoprotein-mimicking nanodisc coupled with antigen peptides and adjuvants markedly improved antigen/adjuvant co-delivery to lymphoid organs and sustained antigen presentation on dendritic cells [21]. Furthermore, advances in nanotechnology may also provide advantages such as higher levels of accuracy during surgical resectioning with technologies such as quantum dots [22] and in situ molecular mapping of cancer biomarkers via use of so-called upconversion nanoparticles [23].

1.2. NCI Alliance for Nanotechnology in Cancer

The National Cancer Institute (NCI) recognizes the need to discover and develop innovative cancer treatment strategies, and has taken notice of the promise these unique nanotechnologies have to offer. In 2004, the NCI Office of Cancer Nanotechnology Research (OCNR) established an integrated, milestone-driven and product-oriented project called the Alliance for Nanotechnology in Cancer (ANC) program [24]. The launch of the NCI ANC program was a groundbreaking initiative for NCI to improve early cancer detection and imaging capabilities and develop novel therapeutic options for cancer patients [25].

Several organizations were formed under the ANC umbrella. The Centers of Cancer Nanotechnology Excellence (CCNE; 2005-2020) initiated multiple projects focusing on integrated technology solutions to significant problems in cancer biology and oncology at both the preclinical and clinical stages [26]. Cancer Nanotechnology Platform Partnerships (CNPP; 2005-2020) were designed to be multi-disciplinary collaborations for the development of unique nanotechnology-based solutions to the major obstacles of cancer biology, treatment and diagnosis. Innovative Research in Cancer Nanotechnology (IRCN; 2005-present) facilitated the fundamental understanding of nanomaterials and nanotechnology’s role in overcoming major hurdles in cancer biology, diagnosis and treatment while simultaneously developing innovative, interactive, and clinically relevant solutions to cancer [27]. Toward Translation of Nanotechnology Cancer Interventions (TTNCIs; 2020-present) focuses on more advanced technologies in the later stages of preclinical development with the goal of transitioning these technologies to other NCI translational programs [28]. Cancer Nanotechnology Training Centers (CNTCs; 2005-2020) were novel research programs designed to train the next-generation workforce to pursue cancer research in a multi-disciplinary and multi-mentor environment. The Nanotechnology Characterization Laboratory (NCL; 2004-present) is a government-funded laboratory dedicated to supporting the extramural research community by developing and standardizing protocols for physicochemical, immunological, pharmacological, and toxicological characterization of nanomaterials and providing free nanoparticle characterization services [29]. Another pivotal product of the ANC was the establishment of caNanoLab, a data sharing and data mining repository for well-characterized, novel, cancer-related nanotechnologies [30].

The NCI Alliance program has established an extensive research network of principal investigators. These scientists conduct cutting-edge research in the pursuit of relief from cancer’s burden and have published over 4600 peer-reviewed research articles in the field of cancer nanotechnology. Moreover, several innovative technologies have come to fruition under the aegis of the NCI ANC. Particle Replication in Non-Wetting Templates (PRINT), for instance, is a technology for large-scale fabrication of nanoparticles with precisely controlled size, shape, surface chemistry, and modulus for various applications in nanomedicine and diagnosis [31]. Another example, the integrated blood barcode chip (IBBC) microfluidic system, performs on-chip plasma separation and in situ protein measurement from microliter quantities of whole blood; the multiplexed detection and quantification of protein biomarkers is achieved with a simple finger prick [32]. Many investigators who benefited from the program have advanced their discoveries from the laboratory through patents and start-up companies. Over the years, ANC members have formed over 100 start-up companies and partnerships eager to bring their efforts from benchtop to bedside [33]. The ANC program’s success has resulted in a convergence of several scientific disciplines, fostering innovation in cancer nanotechnology and hastening translation to the clinic.

1.3. Nanotechnology Characterization Laboratory

The NCL is an NCI-funded resource fostered from the ANC program. The NCI established this contractor-operated laboratory to develop, standardize and disseminate protocols for the thorough characterization of nanomaterials. The NCL provides a free characterization service for cancer nanotechnologies via the so-called Assay Cascade to help facilitate their translation from bench to clinic. This Assay Cascade characterization program provides developers a detailed analysis of the physicochemical, immunology, pharmacology, and toxicology properties of their formulation and assesses the overall translational potential of the formulation. To date, the NCL has characterized more than 460 unique nanomaterials through this program for more than 150 developers. Nineteen NCL collaborators have advanced their products into clinical trials, and four have seen commercialization of their products.

Over the past two decades, the NCL has also worked alongside caNanoLab. The caNanoLab portal was initially designed based on case studies from the NCL, which involved the submission and retrieval of physicochemical, sterility and endotoxin, in vitro, and in vivo properties as defined in the well-established NCL Assay Cascade [34]. In addition to annotating nanomaterials with diverse characterizations, the NCL also shares its protocols and publications on the site, making the caNanoLab database an inclusive site for finding composition, synthesis, characterization and publication information on some of the fields most advanced biomedical nanomaterials.

2. Nanotechnology data mining

2.1. Rise of machine learning and artificial intelligence in biology and medicine applications

Artificial intelligence (AI) is the simulation and imitation of intelligent human behavior by a computer, usually driven by patterns identified in large sets of data. Advances in software and hardware, alongside the development of machine learning and deep learning, have made AI increasingly attractive for biological and medical applications. To illustrate, AI-based computer algorithms have revolutionized interpretation of image-based diagnoses, time series have been employed for electrocardiogram analyses, automatic speech recognition displays high accuracy in the assessment of neurological disorders, 3D context-aware deep learning ranked second place in the testing phase of the 2019 Computational Precision Medicine Radiology-Pathology (CPM-RadPath) Challenge on brain tumor classification, and natural language processing has replaced the need for humans to extract patients’ electronic health record data [35–38]. Some AI algorithms also outperform standard statistical approaches in clinical genomics analysis and genotype-to-phenotype predictions [39–41].

Machine learning involves the “learning” of functional relationships and patterns from past data to make reliable predictions of large, complex, and hard-to-discern patterns of future data. This capability is well-suited for analyses in computational biology, systems biology, genomics, evolution, text data mining, proteomics, and biological imaging [42]. Four steps are involved in a canonical machine learning workflow: data pre-processing, feature extraction, model training, and evaluation [43]. The two main types of machine learning are supervised learning that learns from expert-labeled available training data, and unsupervised learning that learns only general features from unclassified training data [44]. Machine learning techniques have been tools for biomedical data analysis for many years. For example, in 1985, statistical decision theory was employed to weigh the risks and benefits of advanced ovarian cancer treatment [45]. Concomitant with the progression of learning methods and technologies in recent decades, various statistical, probabilistic, and optimization techniques have been realized. Artificial neural networks and decision trees have seen extensive use in susceptibility, recurrence, and survivability predictions of various cancers for almost 40 years [46–61]. In addition, support vector machines create a hyperplane to separate two data sets with maximum margin. Machine learning is a powerful tool for cancer and cancer driver mutation classification, prognosis prediction, anticancer drug selection, biomarker identification, and tumor-specific biological process discovery [44, 62–68].

Deep learning represents a significant advance in machine learning, allowing for the assemblage of raw data into layers of intermediate features [69]. It has seen widespread use in biological data pattern analysis, and various deep learning architectures play major roles in image, signal, and sequence data examination. Deep neural network and convolutional neural network methodologies have been widely used in the integration of pathology images and molecular information [70]. Examples include image segmentation of focal brain pathologies from magnetic resonance imaging [71] and neuronal structures in electron microscope images [72]; malignancy classification through identification of lung nodules from computed tomography scans [73]; detection and decoding of electroencephalography, electrocardiogram and electrooculography signals for pattern classification and diagnosis [74–76]; analysis of expression patterns from gene expression data [77]; reconstructing protein structures [78]; prediction of drug-target binding and microRNA precursors [79, 80]; and creation of transcription factor activity profiles for various cell lines [81].

Together with analyses software, advances in hardware technology have dramatically shaped the practice of medicine. Thanks to rising smartphone usage, AI/smartphone integration now plays a role in numerous medical services. Deep learning algorithms allow the public to access and monitor their healthcare data at any time. In 2017-2018, Apple received clearance from the U.S. FDA for a proprietary Apple Watch Series algorithm that detects atrial fibrillation [82]. In 2019, the FDA approved an AI-based large vessel occlusion stroke platform that constructs a diagnostic and treatment plan for critical stroke patients via text messages [83]. An AiCure platform relies on computer vision algorithms to automate directly observed therapy (Automated DOT^®) via a Health Insurance Portability and Accountability Act (HIPAA)-compliant smartphone application enabling real-time access to a patien′s data, medication dosing history, etc. (NCT02243670) [84].

2.2. Public databases for nanotech data mining

Nanotechnology is a multi-billion dollar market and is only expected to grow as newer technologies continue to come to fruition. Alongside this rapid expansion, a large amount of research data regarding nanomaterial synthesis methods and physicochemical properties has been generated. Additionally, information pertaining to in vitro and in vivo characterizations, activities, and functionalities is growing steadily. Much of this data, however, comes from diverse research fields and is scattered across a wide range of publications. Without an integrated management infrastructure for data extraction, analysis, and sharing within and across disciplines, progress may be hampered and delayed. Over the previous decade, advances in data mining have allowed for extraction of specific informative components from massive data sets. The development of publicly available databases and computational models that coordinate research efforts in a more efficient and reliable way has the potential to expedite and improve nanotechnology research efforts.

To meet the needs of scientists, many nanotechnology databases catering to specific topics have been developed. They are open to the public and contain valuable information generated by the nanoscience research community. Many of these databases have a common feature: they include annotated data on composition, characterization (size, shape, purity, zeta potential, melting/freezing point, cytotoxicity, genotoxicity, immunotoxicity, pharmacokinetics, oxidative stress, etc.), and synthesis methods for a variety of nanomaterial types (Table 1).

Table 1. Summary of Nanotechnology Data Repositories and Resources.

Table adapted from [99, 100].

Site	Content	Web Address
caNanoLab	Cancer nanotechnology research/data sharing; source for nanotechnology protocols and publications	https://cananolab.cancer.gov/
DaNa	Human and environmental toxicology data for nanomaterials	https://nanopartikel.info/en/
eNanoMapper	Toxicology data sharing for engineered nanomaterials	http://www.enanomapper.net
InterNano	Directory of nanotechnology manufacturing services	https://www.internano.org/
ISA-TAB-Nano	Spreadsheet-based format for the exchange of nanomaterial descriptions and characterizations	https://wiki.nci.nih.gov/display/ICR/ISA-TAB-Nano
NanoHub	Cloud-based nanotechnology simulation tools	https://nanohub.org/
Nanomaterial-Biological Interactions Knowledgebase	Repository for nanomaterial synthesis, characterization, biological interactions data	http://nbi.oregonstate.edu/
NanoMILE	Online tool to support nanoparticle design and bioactivity profiles	http://nanomile.eu-vri.eu/
Nanowerk	Collection of nanotechnology news, resources and databases	https://www.nanowerk.com/
NanoCommons Knowledge Base	Computational tools for risk assessment of nanomaterials	https://www.nanocommons.eu/nanocommons-knowledge-base/
OECD	Standardized methods for nanomaterial safety	https://www.oecd.org/science/nanosafety/
PubVINAS	Online modeling tool for nanostructure analysis and visualization	http://www.pubvinas.com/
S2Nano	Predictive nanotoxicity models using curated physicochemical and in vitro cytotoxicity datasets	http://portal.s2nano.org/

The U.S. government has funded several databases over the years. In addition to the ongoing caNanoLab initiative, the Nanomaterial Registry and the Nanoparticle Information Library were early efforts centered around collection of nanoparticle information and data. The Nanomaterial Registry was a National Institute of Health (NIH)-funded database that set standards for data interoperability, ontology, and controlled vocabularies for data sharing. Additionally, it aimed to simplify the process of creating evaluation and similarity criteria for nanomaterial data [85]. The Nanoparticle Information Library (NIL) was developed by the National Institute for Occupational Safety and Health (NIOSH) [86]. NIL provided a platform to identify and prevent health- and safety-related issues associated with nanoparticle synthesis, handling, or usage. Primary sources of this information included the International Council on Nanotechnology Environmental, Health, and Safety, Woodrow Wilson Center for Scholars, InterNano Online Information Clearinghouse, NIOSH Pocket Guide to Chemical Hazards, Registry of Toxic Effects of Chemical Substances, and the Nanomaterial-Biological Interactions Knowledgebase [86]. While the Nanomaterial Registry and NIL are no longer being maintained, the caNanoLab site continues to expand. Resources are being invested to not only grow the availability of annotated datasets, but also to upgrade and enhance the database for improved useability.

Several databases have been developed through European Union funding initiatives as well. eNanoMapper is a computational infrastructure funded by the EU’s Seventh Framework Programme for research, technological development, and demonstration [87]. Besides providing nanomaterial properties, eNanoMapper contains ontology annotation data for nanosafety and toxicological data management of engineered nanomaterials [88]. User-friendly import and export features of eNanoMapper operate on a REpresentational State Transfer (REST) web services application programming interface [89]. NanoMILE was a consortium aimed at understanding nanomaterial-biological interactions. Using a collection of characterization and toxicity data, an online tool, Enalos InSilicoNano Platform, was developed with Novamechanics Ltd. to support nanoparticle design and bioactivity profiles [90].

Several academic institutions have also collaborated to develop nanotechnology-centric databases. PubVINAS was developed by Rutgers University (New Jersey, U.S.) and Guangzhou University (China). It contains annotated nanostructures that are suitable for direct computational modeling and rational nanomaterial design. Developed models are able to accurately predict zeta potential, logP, and cellular uptake properties of new nanomaterials using machine learning and deep learning technology [91]. The Nanomaterial-Biological Interactions Knowledgebase (NBIK) was developed by Oregon State University. NBIK uses zebrafish embryos as a metric to evaluate biocompatibility and toxicity at multiple levels of biological organization (molecular, cellular, or organismal). Results from all nanomaterials studies are visualized on a heatmap [92]. ISA-TAB-Nano is a spreadsheet-based format that provides a common framework based on existing standards developed by the European Bioinformatics Institute [93]. This standard specification enables the submission and exchange of nanomaterials between caNanoLab and NBIK and represents structure-activity-relationships in nanomedicine [94].

In addition to the data mining and simulation tools being developed, databases have also been developed to generally support the nanotechnology researcher by compiling commercially available resources for nanomaterial products and manufacturing, summarizing relevant nanotechnology news and more. For example,. Nanowerk is an online portal that contains information regarding commercially available nanomaterial products [95].

InterNano is a national nanomanufacturing network (NNN) site that not only carries information about NNN conferences and workshops, but also provides a collection of manufacturing services dedicated to nanotechnology [96]. OECD, the Organisation for Economic Co-operation and Development, provides another resource for standards and tools to assess nanotechnology exposure risks and toxicities [97].

Even though various nanotechnology data repositories have been developed which have provided users with information on advanced nanomedicines as the field has matured, challenges remain. Although the development of machine learning platforms to integrate curated databases attracts great interest now, those machine-learning research groups may not yet be fully aware of the potential and needs in the nanotechnology field. In addition, there are multiple types of databases established with different focuses, some lacking “big data” capabilities to capture the field as a whole, and due to their complexity, can be hard to directly translate into training sets [98]. Moreover, there are no standardized metrics for nanomaterial characterizations and data, making comparisons across databases difficult.

2.3. CaNanoLab

With the rapid progress in biomedical nanotechnologies, the NCI Center for Biomedical Informatics and Information Technology (CBIIT), in collaboration with the NCI OCNR, developed a data repository portal–caNanoLab–to capture and share well-characterized nanomaterial data to expedite and validate the use of nanotechnology in the development of cancer research [101]. The caNanoLab portal was initially designed to establish nanoparticle characterization standards in support of data sharing with the CCNEs and caBIG^™ participants [102]. The NCL provided the basic design concept involving the submission and retrieval of characterization information. Since then, the database has evolved to also include synthesis procedures, a plethora of characterization protocols, and a compilation of cancer nanotechnology-focused peer-reviewed publications. As of early 2022, the caNanoLab portal has 1779 nanoparticle samples, 151 protocols, and 2253 publications. Of the nanoparticle samples, 1598 contain physiochemical characterization, 1175 contain in vitro characterization, and 136 contain in vivo characterization.

caNanoLab is based on a nanotechnology information object model (nano-OM) [101]. caNanoLab version 0.5 was released in June 2006; it focused on capturing the NCL workflow and workflow artifacts [102]. Over the years, the database has continued to undergo development and improvement, adding features such as support for characterization metadata, end-user usability enhancements allowing for more efficient data entry and annotation, improved search functions, protocol versioning management, publication cross-reference to PubMed, printing and export functions, and enhanced security. The latest version, caNanoLab version 3.1.0, was released in August 2022 and continues to provide improved usability features, including migration to a cloud-based platform allowing for improved flexibility and scalability.

3. Trends in cancer nanotechnology research

The data collected within caNanoLab and NCL allows for a unique insight of trends and patterns within the nanotechnology space. Users of the caNanoLab data sharing portal have ready access to a plethora of information and can gain an understanding of the most widely used nanotechnology platforms, the various active pharmaceutical ingredients and cancer indications researchers are exploring, which formulations make use of active targeting, the types of characterizations employed, and more. We have collected trends for a variety of these metrics through both the caNanoLab and NCL programs, offering a comprehensive analysis of the most prevalent research activities in the cancer nanotechnology space. We chose caNanoLab and NCL for a deep comparison because of their broad sampling of cancer nanotechnology over a long period, our intimate understanding of their curation, databasing, and evaluation processes, and our direct access to the underlying data. Both caNanoLab and NCL are managed by the NCI, and these two resources serve distinct roles within the research and commercial spheres. This contrast enables us to make meaningful comparisons between these two areas within the common environment of NCI programs.

3.1. Participating institutions

Data can be entered into caNanoLab by any public user; further ANC-funded organizations are required to enter their data. In addition, NCI employs fulltime data curators to comb the literature for unique and advanced technologies. Curators then work with authors to deposit data into the portal. Combined, these entries stem from a variety of sources but most prominently came from academic institutions, which comprised nearly three-quarters of the total entries (73%) (Fig. 1A). The remaining sources of data were from government organizations (10%), medical centers (8%), non-profit organizations (3%), and industry (6%). Of these, 77% were from U.S.-based organizations, and 43% of the entries were from ANC-funded investigators or spin-off companies developed around their funded nanotechnologies.

Fig. 1. Participating institutions.

A breakdown of the types of institutions submitting data to the caNanoLab portal (A) and those submitting nanomaterials to the NCĽs Assay Cascade characterization program (B).

NCL is a competitive, application-based program for characterization of the field’s most promising cancer nanotechnology concepts and is open to researchers from all institution types globally. In contrast to caNanoLab, NCL sees a large portion of materials from industry (42%), with an equal amount from academic institutions (41%) (Fig. 1B). The remaining participating institutions were similar to those found within caNanoLab, with government (11%), medical centers (3%), non-profits organizations (2%) and independent researchers (1%) comprising the remaining field. The international (12%) and ANC-funded (20%) make-up of NCL submissions were about half those seen within caNanoLab.

Both caNanoLab and NCL serve as global resources for nanotechnology data and trends. caNanoLab data is, at present, primarily centered around publications from academic, specifically ANC-funded investigators, while NCL submissions are equally split between industry and academia. caNanoLab has considerably more flexibility for inclusion of data, wherein any thorough datasets (pending approval from the data curators) can be included in the database. NCL, on the other hand, is limited to the several dozen concepts per year which are accepted into the Assay Cascade characterization program and tend to focus on technologies closer to translation. Together, these datasets are presented as an informative assessment of the current landscape of nanotechnology research.

3.2. Nanoparticle platforms

Few articles have provided metrics on the most commonly utilized and investigated nanoparticle platforms. A 2017 article by D’Mello et al. is perhaps one of the most informative, with breakdowns of the various platforms submitted to the U.S. FDA’s Center for Drug Evaluation and Research (CDER) as drug products, spanning the years 1973 to 2015 [103]. As one might expect, liposomes constituted the majority of submissions, comprising roughly one-third of nanoparticle drug submissions. In the most recent years of this analysis (2010-2015), liposomes were the leading nanomaterial drug submission to the FDA (35%), followed by nanocrystals (29%), emulsions (9%) and iron-polymer complexes (8%). Together, these top four platforms constituted over 80% of CDER’s nanotechnology submissions, with more than a dozen additional nanoparticle classifications comprising the rest of the field. These metrics published by the FDA provide a unique insight into the nanoparticle technologies that have undergone regulatory review. The metrics available from caNanoLab and NCL, on the other hand, provide insight into those currently being explored by the research community and potentially offers a glimpse into those technologies the FDA might be seeing in the near future.

More than 1700 nanoparticles are described in caNanoLab. Liposomes constitute only 10% of the samples deposited (Fig. 2A). The most prevalent are metallic nanoparticles (26%) and polymeric nanoparticles (22%). Other technologies include metalloids (silica), quantum dots, carbon-based (nanotubes, fullerenes), dendrimers, emulsions, proteins, nucleic acids, biopolymers, and lipidic nanoparticles. Of note, all metallic-based nanomaterials were grouped together in the analysis of the caNanoLab data, whereas in the FDA study, they were subdivided into several different categories, including drug-metal, metal-protein, metal-nonmetal, and metal-polymer complexes. The sum of these subcategories, however, still amounted to significantly less than the percentage seen in the caNanoLab analysis.

Fig. 2. Nanoparticle platforms.

A breakdown of the types of nanoparticle platforms described in the caNanoLab portal (A) and those submitted to the NCL’s Assay Cascade characterization program (B).

Many of the nanomaterial platforms described in caNanoLab were also reflected in submissions to the NCL. Three of the four top nanomaterial platforms in caNanoLab were also leading the NCL submissions: metallic (22%), liposomes (20%), and polymeric (15%) (Fig. 2B). Micelles constituted 12% of NCL submissions and 3% of the FDA submissions based on the analysis from 2010-2015 [103].

Compared with the trend of submissions to CDER, in which ″first-generation technologies″ such as liposomes, nanocrystals and emulsions dominated, caNanoLab and NCL are suggesting the emergence of polymeric, silica, lipidic, and nucleic acid-based technologies. Polymeric nanoparticles constituted a significant portion of both the caNanoLab and NCL nanomaterials. Based on nanoparticle submission trends observed by the NCL, the use of polymeric nanoparticles has risen over the last few years, aligning with literature wherein several recent reviews suggest a significant clinical potential for this platform due to its versatility with respect to design (e.g., size, molecular weight and hydrophobicity) and biodegradability [104–107]. Likewise, lipidic nanoparticles are also becoming a popular platform choice, owing in large part to the rapid success of lipid nanoparticles used by Pfizer/BioNTech and Moderna for the COVID-19 vaccines Comirnaty and mRNA-1273, respectively. Both companies are, in fact, exploring the use of their lipid nanoparticle technologies in oncology and immunooncology [108]. Nucleic-acid-based technologies, which were not listed in the D’Mello analysis, are being explored as gene therapies, immunotherapies and vaccines [109, 110]. Several recent papers dive into the clinical potential of these therapies and discuss strategies for overcoming translational hurdles [111–113].

3.3. Applications and cancer indications

As might be expected, the major use of these cancer nanotechnologies are for therapeutic/treatment purposes (Fig. 3). About half of technologies within caNanoLab are classified as therapeutics, while 68% of those submitted to the NCL are therapeutics. Nanoparticles used as imaging agents constitute 41% of those in the caNanoLab database and only 16% of NCL submissions. It has been suggested that companies tend to focus efforts on therapeutics over imaging agents as therapeutics tend to have a higher return on investment and lower development costs as compared to imaging agents [114–116]. Other represented nanoparticle applications include multifunctional particles and those used for hyperthermia, photosensitizers, radioenhancers, surgical tools, and more.

Fig. 3. Applications.

A breakdown of the application of nanotechnologies described in the caNanoLab portal (A) and those submitted to the NCL’s Assay Cascade characterization program (B). A more detailed breakdown of the therapeutic uses for nanoparticles within caNanoLab was not easily achieved with the current search indicators.

A more detailed breakdown of the various applications is provided for the NCL submissions. Chemotherapy is the leading application at 39%, followed by targeted therapies (21%) and imaging agents (16%). Together, these three constitute 76% of all NCL submissions. The remaining submissions encompass immunotherapies, vaccines, hyperthermia/thermal ablation technologies, photodynamic/photothermal therapies, radioenhancers/brachytherapy and multifunctional particles. The number of immunotherapy and vaccine submissions, in particular, have been slowly increasing over the last few years, as have the number of chemotherapy and targeted treatments that can be used in conjunction with commercialized immunotherapies such as PD-1/PD-L1 check-point inhibitors. In addition to strategies that directly treat or diagnose cancer, the NCL also accepts nanotechnologies that can be used in any aspect of a cancer patien′s treatment journey. For example, the 4% in the ″other″ category include nanoparticles that can aid in complete surgical resection of tumors, synthetic platelets to supplement patients experiencing low platelet counts, antimicrobials to tune the tumor microbiota for more effective treatment response, and more [117].

In addition to looking at the application of these nanomaterials, we also explored the cancer indications researchers are treating with their technologies. The top five indications represented within caNanoLab were prostate (19%), ovarian (18%), pancreatic (13%), breast (13%) and brain (9%) cancers (Fig. 4A). The top five indications being explored by NCL-submitted nanomaterials were breast (19%), brain (12%), ovarian (13%), lung (11%), and pancreatic (9%) cancers (Fig. 4B). Four of the top five indications were similar for both caNanoLab and NCL, with differences in their order. Cancers in the “other” category were each ≤1%. For caNanoLab these included bladder, cervical and endometrial cancers and myelomas. NCL’s list of other cancers included these, as well as gastric, kidney and liver cancers, lymphoma, mesothelioma and personalized approaches to treating cancer.

Fig. 4. Cancer indications.

A breakdown of the various cancer indications being studied using the nanotechnologies in caNanoLab (A) and the NCL (B). Cancers in the Other category were each ≤1%. Some formulations are being developed to treat multiple cancer indications and thus were included in more than one category.

The most commonly diagnosed cancers globally in 2020 were breast, lung, colon and rectum, prostate, non-melanoma skin, and stomach cancers [118]. Between both caNanoLab and NCL datasets, breast, prostate, and lung cancer were also among the most studied cancers, showing that researchers are investing significant efforts into developing better treatment options for some of the most common cancers. Importantly though, these datasets also suggest significant research efforts are ongoing for notoriously difficult to treat cancers such as pancreatic, ovarian and brain cancers. These cancers do not have specific screening tests and are often diagnosed late, thus offering poorer treatment responses and survival outcomes. For patients with distant or metastatic disease, brain cancers have a 32% 5-year survival rate, ovarian cancers a 30% 5-year survival rate, and pancreatic cancers a dismal 3% 5-year survival rate [119]. Upwards of 75% of pancreatic cancer patients will succumb to the disease within a year of diagnosis [120].

3.4. Active pharmaceutical ingredients

In addition to looking at trends in nanoparticle platforms and cancer indications, the active pharmaceutical ingredients (API) used in the therapeutic nanoparticles were also analyzed. Initially, the API were grouped by drug classification, e.g., anthracyclines, taxanes, camptothecins, etc. Then, the data were plotted for individual drugs, e.g. doxorubicin, paclitaxel, irinotecan, etc. There are some differences in distribution between the caNanoLab and NCL datasets. However, the most frequently used API were taxanes, nucleic acids, anthracyclines, camptothecins and amino acid-based biologics (antibodies, proteins and peptides) (Fig. 5). Individually, DNA was the most represented API in caNanoLab at 20%, followed by doxorubicin (12%), paclitaxel (12%) and docetaxel (8%). For NCL nanomaterials, doxorubicin was the most utilized API (12%), followed by paclitaxel (10%), siRNA (9%), and docetaxel (8%). DNA, which represented one-fifth of caNanoLab API, was only used in 5% of NCL materials, though submission of these materials has been on the rise over the last few years. This analysis also showed that 19% of samples in the caNanoLab database make use of dual API, whereas only 6% of nanomaterials submitted to the NCL have dual API; this may represent the difficulty in translating formulations with multiple API.

Fig. 5. Active pharmaceutical ingredients.

A breakdown of the various active pharmaceutical ingredients used in the nanotechnologies in caNanoLab (A) and the NCL (B). The left graphs show the API by drug category. The right graphs break out the individual drugs used.

These data align with API that are often used clinically. Doxorubicin, paclitaxel and docetaxel are some of the most commonly used chemotherapeutics in the clinic. All three API are used as standard of care treatments for breast cancer, the most commonly diagnosed cancer in the U.S. and globally [118, 121]. However, they all come with associated toxicities—toxicities which researchers hope to reduce through nanoformulation, as has been done for doxorubicin with Doxil and paclitaxel with Abraxane. The prevalence of siRNA and nucleic acids in general speaks to the rising use of these API as well. There are currently four FDA-approved siRNA therapies: Onpattro, Givlaari, Oxlumo, and Leqvio [122], with many others in clinical trials, including many siRNA-containing nanoparticle concepts for a variety of cancer indications [123, 124].

3.5. Targeting

Targeted drug delivery involves modification of the nanoparticle surface to include specific ligands that will interact with specific receptors on the tumor/cellular target, e.g., small molecule targeting ligands on the nanoparticle surface to bind prostate-specific membrane antigen (PSMA) on prostate cancer cells, thereby releasing drug at the intended therapeutic site. There has been a tremendous amount of literature in the cancer nanotechnology field on the promise of nanoparticle-targeted drug delivery to tumors, though no actively-targeted strategies have yet to come to fruition. The majority of nanomaterials depicted in the caNanoLab database and those submitted to the NCL did not employ active targeting strategies and relied solely on passive targeting mechanisms. Nevertheless, active targeting was used in 18% and 22% of nanotechnology strategies for caNanoLab and NCL, respectively. Despite several articles surmising that nanotechnology has over-promised and under-delivered with respect to active targeting abilities [125–127], roughly one-fifth of research activities are still engaging in efforts to improve active targeting strategies in hopes of realizing the full potential of the technology.

Molecularly targeted therapy, which is different from targeted drug delivery, involves the interaction of the API with its intended molecular target [128], e.g., trastuzumab interaction with HER2 and bortezomib interaction with 26S proteasome. Molecularly targeted therapies were used in 4% of nanomaterials in caNanoLab but were higher for NCL submitted nanomaterials (15%). A small population of materials employed both active and molecular targeting strategies.

3.6. Sample characterization

The importance of nanomaterial characterization cannot be understated. Several reviews have discussed the challenges associated with nanomaterial drug development and characterization pitfalls researchers often encounter [129, 130]. Failure to adequately characterize and understand the physical and chemical attributes and the boundaries which impart biological potency has led to the preclinical failures of many strategies. This is precisely one of the reasons the NCL was established, to assist researchers in thoroughly characterizing their nanomedicines. Thorough characterization can promote lot-to-lot reproducibility, which, in turn, helps to safeguard reproducible biological data. This is also a key motivation behind caNanoLab. Every nanomaterial submitted to caNanoLab must have a minimum set of characterization data included. Generally speaking, minimum characterization should include measurement of basic physicochemical properties such as compositional analysis, size/size distribution and zeta potential. Publications reporting only in vitro or in vivo data should also be supported by this minimum dataset, which can be pulled from earlier, related publications or supplied by the corresponding author. The database curator will reject submissions without this minimum characterization data; submissions which include only theoretical values (as opposed to experimental measurements) will also be rejected. This requirement was put in place to help data mining scientists elucidate accurate comparisons and trends between the physicochemical properties and biological data. A list of common sample characterization categories used in caNanoLab employed in NCL’s Assay Cascade is included in Table 2.

Table 2. Routine nanoparticle sample characterization methodologies.

The following are routine characterizations for nanomedicines, based on regulatory requirements and international standards such as ISO and ASTM International. This list is not all-inclusive. Factors such as nanoparticle platform, active pharmaceutical ingredient, excipients, cancer indication and route of administration will all influence the characterization requirements. MTT, 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium; LDH, lactate dehydrogenase.

Sterility	Physicochemical Characterization	In Vitro Characterization	In Vivo Characterization
β-Glucans	Composition	Autophagic dysfunction	Absorption, distribution, metabolism & excretion (ADME)
Endotoxin	Lot reproducibility	Caspase 3/7 activation	Adjuvanticity
Microbial contamination	Molecular weight & polydispersity	Chemotaxis	Efficacy
Mycoplasma contamination	Nanoparticle concentration	Complement activation	Local lymph node assay (LLNA)/local lymph node proliferation test (LLNP)
	Purity	Cytokines, chemokines & interferons	Pharmacokinetics
	Shape/morphology	Cytotoxicity (MTT & LDH release)	Rabbit pyrogen test
	Single particle characterization	Drug release	Single- and repeat-dose toxicity
	Size/size distribution	Glutathione	T-cell dependent antibody response (TDAR)
	Stability	Hemolysis
	Solubility	Human lymphocyte activation
	Starting material characterization (e.g., drug epimerization)	Leukocyte procoagulant activity
	Surface characterization (e.g., quantitation of surface ligands)	Leukocyte proliferation
	Zeta potential	Lipid peroxidation
		Nitric oxide production
		Phagocytosis
		Plasma coagulation
		Platelet aggregation
		Reactive oxygen species

The routine sample characterizations entered into caNanoLab are depicted in Fig. 7A and 1C. More than half of all characterizations entered are physicochemical assays (55%), roughly one-third are in vitro assays (37%), and small populations are assays for sterility (3%) and in vivo studies (5%). The majority of the physicochemical properties were measurements for size (e.g., by dynamic light scattering or electron microscopy) and surface properties (e.g., zeta potential), comprising 83% of the total physicochemical characterizations entered. Within the in vitro assays, cytotoxicity studies comprised nearly one-half of all entries (49%) and active targeting studies nearly one-third (34%). The in vivo studies were nearly equally split between pharmacokinetics (55%) and toxicology (45%).

Fig. 7. Sample characterization.

A breakdown of the types of characterization conducted on nanomaterials in caNanoLab (A) and the NCL (B). The specific types of assays conducted for physicochemical, in vitro, and in vivo characterizations are also summarized for caNanoLab (C) and NCL (D).

The sample characterization trends for NCL are shown in Fig. 7B and 7D. In vitro characterization constituted just under one-half of all assays conducted (45%), physicochemical techniques were about one-third (35%), sterility assays were 13%, and in vivo studies were 5%. NCL studies also included small populations of ex vivo characterization and modeling studies. Size, composition and surface (primarily zeta potential) constituted over three-quarters of the physicochemical characterization assays in the NCL Assay Cascade. Other techniques included shape/morphology, purity, stability, molecular weight and particle concentration. Within the in vitro studies conducted, hematology, cytotoxicity, and immunotoxicity comprised 84% of all studies, with in vitro drug release and other immunology, general toxicology and pharmacology assays comprising the remaining characterization data. The in vivo studies were equally split between pharmacokinetics (27%), toxicology (26%), and efficacy (23%) as the top three types of studies conducted at the NCL; dose-range finding studies, immunotoxicity, imaging and other studies constituted the remaining one-quarter of the in vivo studies.

These graphs represent a snapshot of all characterizations entered into caNanoLab and those conducted on NCL submissions. They do not, however, show the emerging trends in characterization over time. For example, at NCL, size characterization has shifted from simple batch-mode DLS measurements to flow-mode DLS measurements made after particle incubation in human plasma to provide the most physiologically relevant size possible. In vitro characterization has expanded from routine cytotoxicity and hematology to a variety of mechanistic immunotoxicology assays and comprehensive in vitro pharmacokinetic studies to quantify nanomedicine drug release. Both the number and types of in vivo studies conducted have also increased over the years and now include studies for adjuvanticity, autoimmunity, T-cell dependent antibody responses, and more.

3.7. Protocols

Both caNanoLab and NCL serve as great resources for protocols related to nanoparticle characterization and sample preparation. In fact, part of the NCL′s charter is to develop and standardize characterization protocols in the areas of physicochemical characterization, immunology, pharmacology, toxicology, and sterility. To date, the NCL has over 70 protocols, all of which are made freely available to the research community on their website [131]. These protocols serve as the foundation to NCL′s Assay Cascade characterization program and average more than 2200 downloads every year. They were developed based on scientific justification in the context of regulatory requirements for drug safety. A breakdown of the NCL protocols by category is included in Fig. 8B and includes assays in the areas of physicochemical characterization (PCC), immunotoxicology assays (ITA), in vitro efficacy assays (IEA), general toxicology assays (GTA), pharmacology (PHA) and sterility (STE). It is worth noting that the NCL methodology portfolio includes more protocols than those posted on the website. The protocols released on the laboratory website represent those standardized for a wide variety of nanomaterials. However, many nanomaterials require a customized approach; this leads to the development of assays tailored for the characterization of a specific formulation. Many physicochemical characterization techniques and all in vivo studies are individually tailored for each nanomedicine product and developed on a case-by-case basis. To share this broader characterization knowledge with the extramural research community, the NCL releases guides for materials characterization. There are currently five such guides: three for physicochemical characterization, one for endotoxin, and one for in vivo pharmacokinetic, efficacy, and toxicity study design.

Fig. 8. Protocols.

A breakdown of protocols available in caNanoLab (A) and the NCL (B). PCC: physicochemical characterization; ITA: immunotoxicity assays; IEA: in vitro efficacy assays; GTA: general toxicity assays; PHA: pharmacology assays; STE: sterility assays.

The consistency in applying this standardized portfolio of protocols to the characterization of various nanomaterials allows NCL to identify trends in biocompatibility and link these trends to physicochemical properties. For example, the NCL immunology assay cascade frequently detects chemokine responses and prolongation of plasma coagulation time [132]. A close look at the materials’ composition reveals that most of these materials are either lipid- or polymer-based or contain lipids and polymers as excipients. As a result, nanoparticle-specific protocols to measure the complete composition (for example, lipid, polymer, drug, coating, targeting ligand, and excipient concentrations to name a few) for a given formulation are constantly in demand. This is a prime example highlighting the need for robust characterization—beyond simple size measurements—to better understand the new and often complex, emerging nanoformulations and how their physicochemical properties influence biocompatibility. Similarly, the NCL has sought to develop bioanalytical methods that are applicable to a wide range of nanomedicines, and can be used to compare pharmacokinetics across and between platforms. An example would be the popular stable isotope ultrafiltration assay (SITUA) that has been used to characterize in vitro drug release and in vivo pharmacokinetics of encapsulated and unencapsulated drug fractions for diverse nanomedicine platforms and API [133]. Future standardized NCL protocols in the bioanalytical area are focusing on high resolution determination of nanomedicine distribution in tissue sections, using immunohistochemistry labels [134, 135] and development of discriminatory methods to evaluate total dosage form drug release for solubilizing nanomedicines. It is anticipated that making these data available for mining by bioinformatics researchers will reveal even more trends not easily identifiable otherwise.

Among the 130 protocols in caNanoLab, the majority (69%) are from NCL’s standardized analytical cascade (Fig. 8A). The remaining 31% of protocols are non-NCL Assay Cascade protocols that extramural researchers have developed. These include techniques such as sample preparation of various nanoparticles, in vitro transfection, and in vivo imaging protocols. Additionally, 5% of the protocols on caNanoLab are video protocols, including videos depicting gold nanoparticle synthesis, in vitro cytotoxicity and transfection assays, and in vivo tumor imaging. Furthermore, the caNanoLab database maintains all previous versions of any protocols deposited onto the site. While the NCL website contains only the most current versions of their Assay Cascade protocols, the versioning system within caNanoLab allows users to find exact procedures that may have been used in the previous characterization and/or publication data entered into the portal.

3.8. Publications

A search of the PubMed database (https://pubmed.ncbi.nlm.nih.gov/) using the search terms Cancer AND Nanomedicine OR Nanoparticle OR Nanotechnology yields over 350,000 publications spanning more than 40 years. Without the proper keywords and metadata, further searching and refining of these publications can be time-consuming, especially in the context of data mining; extraction of characterization information and data for a trend analysis would require significant resources. To date, there are 2253 publications entered into the caNanoLab database. While this is only a small fraction of the available literature, the inclusion of searchable raw data makes the caNanoLab subset of publications ideal for data mining scientists. Researchers can quickly search the database for publications associated with sample data, searching by nanomaterial type, functionalizing entity or function. In addition, users can search and extract publications with specified datasets. For example, a user can search for liposomal samples with doxorubicin, used as a chemotherapeutic, and containing a minimum characterization set of size, composition, in vitro hematology, and in vivo pharmacokinetics.

There are, however, limitations to the current version of the caNanoLab database. First, since these publications are manually entered, not only is the number of publications significantly lower as compared to those available in PubMed, they are also entered at a slower rate. Investigators cannot rely solely on caNanoLab for a full spectrum of available cancer nanotechnology literature, at least not at the present time. Second, the majority of publications within caNanoLab were published only in 2006 and later. While this cannot provide a look at earlier research (at least at present), it does limit the publications to those with a minimum characterization dataset. This helps to ensure that in vitro and in vivo biological data are representative of the nanoparticle’s stated physicochemical properties. Lastly, a significant portion of the publications have yet to include the raw data. While datasets are incorporated as graphs and figures, researchers have been reluctant to provide the necessary raw data to meet the requirements for data mining. Though, as artificial intelligence, data mining and machine learning technologies are quickly coming to the forefront of biomedical research, investigators are more likely to realize the importance of this contribution.

4. Vision for future research

Cancer nanotechnology has been advancing for decades. Progress has been made for cancer diagnosis, therapy, drug design and delivery, and clinical achievements have been realized for a variety of drugs. Despite these advancements, there is still ″Plenty of Room at the Bottom″ [136]. Reliable cancer biomarkers capable of identifying signatures of tumorigenesis, progression, metastasis, and monitoring prognosis are urgently needed. Surface engineering of nanoparticles can allow multiple ligands to be readily incorporated, generating a multiplexed molecular profile of upregulated tumor targets for improved therapeutic targeting and treatment. Attachment of functional groups with various optical, radioisotopic, photoacoustic, or magnetic properties can also improve the efficiency and accuracy of in vivo molecular imaging.

For several decades, the efforts of cancer nanomedicine developers focused on masking nanocarriers from immune recognition. However, recently it became evident that the inherent property of the immune system to recognize and clear particulate materials, regardless of their stealth surface modifications, has collateral benefits that can aid the efficacy of anti-cancer treatments. This immune cell targeting can be thought of as a form of passive targeting, as it depends on the physicochemical properties of nanomaterials, as opposed to active targeting ligands. This natural immune system tropism of nanoparticles has been embraced for delivery of immunomodulatory and vaccine nanomedicines [137]. More and more, recent studies describe adjuvant properties of drug-free nanocarriers that, through various and not always understood mechanisms, activate antigen-presenting cells and promote specific T-cell responses. For example, a recent study demonstrated that lipid nanoparticles used for delivery of mRNA encoding SARS-CoV-2 S-protein antigen induce T follicular helper cells and promote humoral responses through a previously unknown adjuvant mechanism [138]. Immunomodulatory nanoparticles are increasingly investigated for optimizing the tumor microenvironment to make it favorable for checkpoint inhibitor therapy [139–141]. Approved cancer nanomedicines, such as Doxil, demonstrated a greater rate of complete anti-tumor response in combination with checkpoint inhibitors [142]. Although the induction of the so-called immunogenic cell death by cytotoxic drugs and nanocarriers was proposed as a potential mechanism [143, 144], it alone cannot explain the observed improved efficacy. Therefore, understanding the immunological properties of nanocarriers became more crucial than ever for designing safe and efficacious therapies. It is intuitive that applying machine learning and developing AI algorithms to the growing multi-parameter nanoparticle characterization datasets will further streamline the design and selection of nanocarriers with desirable immunological properties. Some such studies are already underway. For example, several earlier studies investigated the cytokine response to nucleic acid nanoparticles (NANPs) of various shapes and sizes [145, 146]. Next, machine learning and AI algorithms, including the modified transformer algorithm, were applied to the analysis of these data sets and resulted in a publicly available online tool (https://aicell.ncats.io/) for the selection of oligonucleotides that, once assembled into a NANP, would produce an expected (low or high) level of interferon responses [147]. Another study applied machine learning to develop a lightGBM algorithm for the analysis of substructures of ionizable lipids with the purpose of identifying critical components necessary for LNP-mRNA formulation; the authors proposed using this algorithm to design LNP-mRNA vaccines [148]. More studies are needed to optimize the existing models, promote their use in nanomedicine research and validate across various types of nanomaterials and applications.

One area of scientific inquiry and regulatory scrutiny that has always plagued nanomedicine research is pharmacokinetic and biodistribution evaluation. Fundamental to the concept of nanomedicines is the ability for nanoparticles to influence drug delivery and distribution, but nanomedicine pharmacokinetic research is complicated by the importance of evaluating drug fractions, i.e. nanoparticle encapsulated and unencapsulated drug, both in the systemic circulation and tissues. Understanding of encapsulated and unencapsulated drug concentrations is vital to optimize drug properties and treatment regimens, and relate pharmacokinetics to pharmacodynamics, as well as comparison of complex generic drug similarity [149]. While there has been substantial progress in development of novel methods to evaluate nanomedicine drug fraction systemically, the ability to measure drug fractions in tissue has lagged and is an area of ongoing research [150]. Recent analysis of nanomedicine pharmacokinetic databases for trends in tumor distribution came to disparate conclusions, with a 100-fold difference in the calculated tumor uptake efficiencies [151]. It is noteworthy that both surveys relied on total drug pharmacokinetics due to the lack of availability of drug fraction pharmacokinetic data, and researchers noted that, if available, drug fraction pharmacokinetic data would have greatly enhanced their analysis and conclusions.

Data mining, AI, and machine learning will be critical in advancing the nanomedicine field and expediting translation. Though many of these efforts are currently aimed at analyzing medical images to compensate for shortages of qualified radiologists and pathologists [152], the future of these computer technologies is not limited to clinical practice, and earlier bioinformatics efforts can share lessons helpful to current cancer nanotechnology data-sharing initiatives. For example, AlphaFold is an artificial intelligence program that performs protein structure predictions [153]. In July 2022, AlphaFold released the predicted structures of more than 200 million proteins from 1 million species, covering nearly every known protein on the planet [154]. Drug discovery, on the other hand, has evolved from “random searches” [155, 156] to high-throughput screenings, but with extremely low hit rates of 0–0.01% [157]. On top of the low hit rate, clinical successes are further reduced by complicated requirements for designing and optimizing appropriate drug delivery platforms. The new era of artificial intelligence and machine learning hopes to shorten this time to discovery and increase overall success rates [158]. Although there have been significant advances, researchers must rely on mining high-quality, big datasets with accurate models to achieve this realization. According to the so-called 80/20 principle, 80% of a data scientist’s time is spent collecting, organizing, and cleaning the data, while only 20% is spent on analyzing and processing the data [159]. Learning models that mimic human intelligence with high accuracy are in great demand. Despite the numerous challenges and drawbacks data mining, artificial intelligence and machine learning still face, these technologies offer the scientific community invaluable benefits, especially in age of personalized/precision medicine, which has the potential to revolutionize medicine and clinical practice [160].

Fig. 6. Targeting.

A breakdown of nanotechnologies using active targeting and molecularly targeted treatment strategies in caNanoLab (A) and the NCL (B).