Compute and interpret BY ipSAE, ipTM, and composite scores from PAE matrices.
npx skills add https://github.com/001TMF/blatant-why --skill by-scoring
Please help me install this Agent Skill. Skill: by-scoring Source: https://github.com/001TMF/blatant-why/tree/main/templates/.claude/skills/by-scoring Command: npx skills add https://github.com/001TMF/blatant-why --skill by-scoring
BY Scoring interprets and applies BY's custom scoring metrics — ipSAE, ipTM, pLDDT, RMSD, liability counts, and the composite ranking formula — to rank antibody, nanobody, and de novo designs based on interface confidence, global topology, and manufacturability.
Run the BY scoring workflow on a Protenix output to produce ipsae_min, iptm, and a composite ranking.
High-intent search queries and answers about installing and using this skill.
BY scoring computes ipSAE and ipTM scores from a PAE matrix and chain assignments to evaluate interface confidence, producing per-design metrics and a composite ranking for antibody and nanobody designs.
The BY composite score ranks de novo designs by combining ipSAE_min, ipTM, pLDDT_mean, RMSD, and liability counts into a unified metric reflecting interface confidence, global topology, and manufacturability for final verdicts.
Apply multi-seed scoring and batch ranking by processing PAE matrices across seeds with hard modality filters, generating per-design ipSAE, ipTM, composite scores, ranks, and verdicts for candidate panels.
Yes, BY scoring processes Protenix outputs by interpreting PAE matrices to score designs after refolding, calculating ipsae_min, iptm, pLDDT_mean, and composite rankings to guide candidate selection decisions.
Metric disagreements arise because ipSAE measures interface confidence while ipTM captures global topology; BY scoring resolves this by computing a liability-based composite score and final verdict to rank designs consistently.
alex-hhScore predicted protein-protein interactions using ipSAE, pDockQ, pDockQ2, and LIS metrics.
solo-istScore PR review findings with a quantitative rubric and output routing decisions.
001TMFDesign antibody, nanobody, and de novo binder backbones conditioned on target structures.
ANakaAnalyze AlphaFold/ESMFold predictions and visualize pLDDT and pAE confidence metrics in PyMOL.
