Generalized Read-Across (GenRA)
Chemical read-across is a commonly used technique to fill data gaps. Existing data for one substance (the source) is used to predict the same endpoint(s) for another substance (the target) that is lacking data and considered to be ‘similar’ in some way (e.g., structural similarity). Read-across is a subjective, expert-driven approach which is challenging from a reproducibility and scalability perspective. GenRA is an algorithmic approach to permit objective and reproducible read-across predictions of in vivo toxicity and in vitro bioactivity.
Available as a web-based application, GenRA can be accessed directly or from within EPA’s CompTox Chemicals Dashboard following a chemical search. Since the applications are linked, users are able to access the wealth of information disseminated within the Dashboard before making a determination that a read-across approach is merited using GenRA. GenRA is structured to mirror the read-across workflow typically used in expert assessments by which candidate source analogues are first identified, evaluated, and then data gap filling is performed using GenRA’s similarity weighted activity approach. Structural and bioactivity information or a combination of both can be used to identify candidate source analogues with available in vitro bioactivity and/or in vivo toxicity data.
GenRA provides many functionalities, including the ability for users to:
- Search for candidate analogues based on different identifiers or draw a chemical structure using the Ketcher drawing pad;
- Identify candidate source analogues on the basis of chemical and bioactivity fingerprints or a combination of fingerprints using the custom hybrid option;
- Download the top 100 candidate source analogues, their pairwise similarities and their chemical/bioactivity fingerprint matrices;
- View chemical neighborhoods via a neighborhood explorer graph visualization tool without filtering on the basis of ToxCast or ToxRefDB data;
- Compare the distribution of relevant physicochemical properties across candidate source analogues;
- Make binary in vitro predictions of ToxCast assay outcomes or binary in vivo toxicity predictions on the basis of study type-toxicity effect using ToxRefDB data;
- Make potency-based predictions of study type-toxicity effects using ToxRefDB data;
- Sort predictions based on number of positive/negative toxicity effects or on the basis of the prediction confidence and download the predictions as a spreadsheet report.
Web Application
Publications and Resources
- Towards systematic read-across using Generalised Read-Across (GenRA) (2023 publication)
- GenRA Manual