Making Assays FAIR

Overcoming Barriers to the Adoption of Human Alternatives to Animals in Biomedical Research

FAIR data is data that is findable, accessible, interoperable, and reusable. Well, maybe it’s time for FAIR assays. Human cell-based in vitro assays and platforms (in vitro NAMs) are promising alternatives to animal testing in biomedical research.

Making in vitro NAMs findable, accessible, interpretable and their data reusable will help drug discovery researchers and regulatory scientists understand and gain confidence in using these tools for decision-making.

The Evolution of Biomedical Research

The last 3 decades have seen a dramatic shift in how we conduct biomedical research. When I was starting out, experiments were small, narrowly focused on answering an individual hypothesis-driven research question. Our knowledge of human biology moved forward in incremental steps.

Biomedical research is done differently now that technology advances have unlocked experimentation at scale. Instead of addressing one question at a time, 100s of questions can be answered with a single experiment.

High throughput screening (HTS) campaigns of large compound libraries were one of the first areas to benefit from scalability. The success of HTS drove further advances in automation and adoption of production level infrastructure in laboratories. Efforts quickly extended beyond biochemical assays to cell-based assays. Now automated platforms have become quite sophisticated, able to accommodate long term cell culture, complicated treatment schedules, as well as manage image- and time-based outputs.

Automation and engineering advances have also made it possible to iterate experiments at scale. Rapid improvements in cell culture methods have led to the subsequent explosion in the numbers and types of human in vitro cell-based assays and platforms, so-called novel alternative methods or in vitro NAMs. These methods include a variety of formats, from static 2D cultures to organoids and tissue chips or microphysiological systems (MPS).

The applications of human in vitro NAMs in biomedical research are broad ranging and include target identification, mechanism of action characterization, biomarker discovery, safety evaluation and assessment of pharmaceutical properties. Despite their potential, adoption of these tools in biomedical research to reduce and refine the use of animals has lagged. This has spurred interest in “catalyzing the development and use of novel alternative methods to advance biomedical research”. See recent announcements from the NIH (NOT-OD-23-140) and FDA (FDABAA-24-00123).

And despite the number and variety of in vitro NAMs that have already been developed, they can be surprisingly hard to find. The provider landscape is diverse and fragmented. In vitro NAMs aren’t consistently described, and no central assay registry exists. They can be complicated and difficult to interpret. Data sets have inconsistent formats, feature different analysis methods, and are siloed.

Efforts to make in vitro NAMs assays “FAIR” (Findable, Accessible, Interpretable and Reusable) should help build scientific confidence in these approaches and catalyze their use in biomedical research to bring in more human-relevant science.

FAIR Assay Principles

FAIR assay principles (findable, accessible, interpretable, reusable) borrow from the well-known FAIR data principles in data governance (findable, accessible, interoperable, and reusable). See Wilkinson et al, 2016 or visit Go-FAIR.org.

Findability

Despite the number and diversity of in vitro NAMs available, they are not easy to find. The terminology to describe assays isn’t well standardized and on the commercial side, there are many vendors. While collections of “protocols” and data sets abound (see PubChem, ChEMBL, and Protocols.io), these are not consistently curated, and lack harmonized standards.

There have been efforts to describe assays using standardized metadata fields (Wang et al., 2017; Daniel, 2022; Vanderwall, 2023). And some organizations have developed proprietary standards for their own internal use. Developing a common standard for describing in vitro NAMs and a centralized registry of available assays would help make these tools more visible to researchers and regulators.

Accessibility

In vitro NAMs are complex and can be challenging to run consistently. Even relatively simple assays contain dozens of components that impact the quality of assay results (cell source, cell state, media, reagent lots, timing, etc.). Consistent and reproducible assays require significant infrastructure and good laboratory practices to ensure quality performance.

Few labs have the sufficient infrastructure to run production level assay operations. Centralization of operations helps reduce costs, provides higher quality results, and should be leveraged to make platforms available more equitably to a wide range of researchers. In addition to centralized core facilities, supporting the commercial ecosystem around in vitro NAMs will be important for promoting wider adoption of these tools and speeding up the advancement of biomedical research.

Interpretability

The complexity of in vitro NAMs and diversity of assay formats can make them hard to understand. If we’d like to see greater use of these tools to reduce or replace animal studies, researchers will need better and more consistent access to reference data with greater standardization around their characterization and analysis methods.

Standardizing how assays are characterized for performance and making characterization data available in user-friendly formats can help. Aligning on common characterization standards provides a minimum threshold for encouraging quality data and would promote good assay development practices. It would help support assessment of technology readiness and the development of robust data sets required for formal validation efforts.

There are two aspects of assay characterization: technical (performance metrics) and biological (relevance and scope). Performance metrics include measures of response (e.g., signal to background) and variability (e.g., %CV, Z’-factor), important metrics for assessing reproducibility. Biological scope covers the system components: cell types and constituent proteins, metabolites, and morphologic features within the assay; but more importantly, the range of functional responses.

Positive controls and reference agents (drugs, known chemicals, etc.) are critically important for defining the range of functional responses and characterizing the biological scope of an assay system. Aligning on key positive controls will be an important goal for community efforts. Beyond references lists (see here, here, and here), collating and sharing detailed information on individual reference agents (e.g., appropriate concentrations, limitations, activities in other assays, literature references, etc.), information that supports their selection as a positive control will help assay users make more accurate conclusions about assay results and the relevance of these results for a specific context of use.

Cost factors will limit what can be included in a minimal set of characterization standards. However, additional characterization criteria could be developed over time, for example, global or single cell transcriptomics, metabolomics, or proteomics; responses to defined sets of nuisance compounds or to expanded sets of reference agents.

Reusability

One of the biggest opportunities for in vitro NAMs to advance biomedical knowledge comes from the ability of scaled platforms to generate large reference data sets. Such data have been used in predictive models, for identifying novel target, mechanisms of action and toxicity hazards (Akbarzadeh, 2022; Berg, 2021; Subramanian, 2017). These data sets support precision medicine, contribute to biomedical knowledge graphs and represent valuable inputs for other approaches such as large language models (LLMs). Integration of these data with other sources requires that the data be effectively organized, structured, and annotated. Efforts to create registries, harmonize nomenclature, develop ontologies, and support data integration are important for data reusability. In addition, training of biomedical researchers in understanding best practices for experimental and data management an improving the level of data literacy amongst biomedical researchers will be needed.

Projects that Support FAIR Assays

Numerous organizations already support projects and initiatives that support FAIR assays. Government research centers such as EURL ECVAM at OECD and NICEATM at the NIEHS are involved in projects that support validation of NAMs for regulatory toxicity testing. The OECD’s Guidance Document on Good In Vitro Method Practices (GIVIMP) describes assay performance best practices and quality considerations. Many organizations and academic groups have contributed to biomedical ontology projects (see BioPortal for their repository of >1000 ontologies).

Research software companies such as Collaborative Drug Discovery are developing tools that support assay registration with metadata annotation using public ontologies. The Pistoia Alliance supports several efforts in support of standardized assay metadata models including the Data FAIRy Bioassay Annotationproject and the Pistoia Alliance/FDA In Vitro Pharmacology Community.

We ourselves are working on a project to catalog human in vitro NAMs to promote greater use of human-based alternatives to animals. This catalog, CHITA, Catalog of Human In Vitro Translational Assays, will be an open-access, curated registry and data portal.

The CHITA catalog will contain information on human cell-based assays and platforms from diverse assay providers. The goal is to assist drug developers and regulatory scientists in finding and gaining confidence in human based in vitro NAMs. The CHITA platform will unify standardized assay metadata and assay characterization data sets in a user-friendly database portal, with search and visualization tools aimed to serve biomedical research biology leaders.

We hope that such a catalog can address some of the barriers to adoption – the lack of common understanding about various platforms, their capabilities, and limitations. To participate or find out more, contact us here.

Photo by Daniel Roe on Unsplash.