CertisAI looks at molecular features of drugs, rather than drug targets. The stated targets for these drugs are not always the only ones present – off-target features can be present too. Similar to how a chemist looks at a compound, the CertisAI model recognizes certain molecular fingerprints in drugs, and then can predict efficacy in PDX and other cancer models. CertisAI has the potential to learn predictive features and molecular profiles, too.