peptide-discovery

Explore and annotate your sequences interactively, lets you write rule-based annotations that tag sequences by any criteria you choose

Compute learned per-sequence vector representations for peptide and antibody/TCR datasets. Runs protein language models over your dataset so you can compare, cluster, and map a repertoire by learned similarity

Group sequences by learned similarity rather than shared letters. Clusters clonotypes or peptides by their distance in protein language model embedding space using HDBSCAN, finding related sequences that identity-based clustering separates because their amino acids differ.

Imports assay data — binding measurements, antigen specificity, functional readouts — and uses MMseqs2 to align its sequences against your clonotypes or peptides, so the measurements land on the right candidates even when the sequences are not byte-identical.

Compares clonotype or peptide frequencies across the conditions of a selection experiment — phage display rounds, yeast display sorts, in vivo timepoints — scores how strongly each sequence enriched, and ranks the candidates that came up.

Compute physicochemical properties (charge, GRAVY, pI, MW, ε) for peptide and antibody/TCR lead ranking

Groups similar sequences (clonotypes or peptides) into clusters based on similarity using MMseqs2.