Traditional read-across is expert driven and involves subjective selection of analogues and data interpretation.

Propose generalized read-across, called GenRA, to predict similarity-weighted toxicity of chemicals in local domains.

GenRA uses chemical structure and ToxCast bioactivity measure similarity.

Evaluate GenRA toxicity predictions objectively for local domains using receiver operating characteristic (ROC) curve.

Compare predictive performance of GenRA using chemical and bioactivity descriptors across local domains.