Every FDA-approved targeted cancer drug has a resistance landscape — the set of single-residue mutations that, under selection pressure, will break it. The atlas surfaces the top predicted resistance events per drug, triangulated from rigid-receptor Δ-scoring, ESM2 protein-language-model fitness, codon-mutational accessibility, and a literature prior. Every prediction is timestamped, citation-backed, and publicly re-derivable from open code. Pre-registered forecasts compound the credibility ledger with every clinical confirmation that follows.
Atlas v1 (May 2026) backfills published clinical resistance events as a calibration anchor. v2 onward will surface NOVEL predictions — residues the model flags that have NOT yet appeared in clinical literature — with the date stamped at first publication.