Represents one paired antibody heavy- and light-chain sequence as 480 numerical features for sequence comparison and downstream analysis. These features do not predict structure, binding affinity, or efficacy.
Represents one protein sequence as 480 numerical features for downstream classification and comparison, not structural coordinates, binding affinity or functional scores.
Computes Morgan-fingerprint Tanimoto similarity between one reference molecule and up to 1,000 candidates on a GPU. Similarity compares structure; it does not indicate binding or potency.
Small-molecule modelingStructure search
Main input data
File
Original usage terms · Commercial use allowed · check notices
Generates sequence and all-atom structure together while keeping the design conditions written in an annotated CIF. One run produces one to four candidates and no quality score.
Protein designEnzyme designMotif scaffolding
Main input data
File
Original usage terms · Commercial use allowed · check notices
Scores how plausible each mutation looks to a language model that reads sequence together with structure. This is evolutionary plausibility, not a stability ΔΔG.