POISK co-folding binder-versus-decoy benchmark
Scores structure prediction models as zero-shot binder classifiers on 22,494 antibody-antigen complexes drawn from patent filings, by how their interface confidence ranks patent-supported binders against putative non-binders for the same antigen.
Results are recorded on the linked protocols, tasks or configurations listed below. Their settings and evaluations are kept separate.
Overview
An explanatory profile has not yet been reviewed for this record.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
Results on linked records
This benchmark has no results of its own in this release. 13 linked records have results. They are not assigned to the underlying benchmark: each record keeps its own settings and evaluations, and its scores are on its own page.
- POISK binder-decoy separation in full-Fv ipSAE, Cohen's d Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK binder-decoy separation in ipTM, Cohen's d Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pair coverage under cross-model structural consensus Protocol · part of this benchmark · 2 evaluations · 2 results
- POISK pairwise binder-versus-decoy accuracy under cross-model structural consensus Protocol · part of this benchmark · 2 evaluations · 2 results
- POISK epitope contact recovery Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pairwise binder-versus-decoy accuracy at 10% coverage Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pairwise binder-versus-decoy accuracy at 100% coverage Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pairwise binder-versus-decoy accuracy at 25% coverage Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pairwise binder-versus-decoy accuracy at 50% coverage Protocol · part of this benchmark · 6 evaluations · 12 results
- POISK pairwise binder-versus-decoy accuracy at 75% coverage Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK pairs left undecidable by identical ipSAE Protocol · part of this benchmark · 6 evaluations · 6 results
- POISK 1-in-50 screening simulation, top-1 retrieval Protocol · part of this benchmark · 6 evaluations · 12 results
- POISK 1-in-50 screening simulation, top-10 retrieval Protocol · part of this benchmark · 2 evaluations · 2 results
Methods and evaluation design
Procedure, tasks and evaluated configurations
Additional methodological details, tasks and evaluated configurations have not yet been extracted into a structured profile for this benchmark. Consult the linked sources and evaluation records below.
Baseline coverage
Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.
0 of 26 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.
Baseline status by linked protocol
- POISK binder-decoy separation in full-Fv ipSAE, Cohen's d · 0/2 roles measured
- POISK binder-decoy separation in ipTM, Cohen's d · 0/2 roles measured
- POISK pair coverage under cross-model structural consensus · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy under cross-model structural consensus · 0/2 roles measured
- POISK epitope contact recovery · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy at 10% coverage · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy at 100% coverage · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy at 25% coverage · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy at 50% coverage · 0/2 roles measured
- POISK pairwise binder-versus-decoy accuracy at 75% coverage · 0/2 roles measured
- POISK pairs left undecidable by identical ipSAE · 0/2 roles measured
- POISK 1-in-50 screening simulation, top-1 retrieval · 0/2 roles measured
- POISK 1-in-50 screening simulation, top-10 retrieval · 0/2 roles measured
Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)
Coverage is derived from release 2026-11-21-c8bea25e5abb. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run instructions
No runnable recipe has been reviewed for this benchmark. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
Release 2026-11-21-c8bea25e5abb · Record review: source checked
2 source records and release history
- Tupikina et al. 2026, Can Folding Models Tell Binders from Bluffers? Evidence from POISK (bioRxiv v1 PDF) · Original source · v1, posted 26 October 2026
- POISK dataset version 1.1 on Zenodo (record 21775623) · Original source · 1.1, published 23 October 2026
Technical metadata and record identifiers
Technical metadata for this record is tracked under stable ID poisk-20261115-benchmark-binder-decoy.
Source locator: Sections 2.2, 2.3, 3.2 and 3.3; Appendices C, D, F and H
Related records
- evaluates task: Discriminating antibody binders from sequence-similar putative non-binders
- uses data: POISK co-folding binder-decoy benchmark set (22,494 complexes)
- subject: POISK: affinity correlation
- subject: POISK: compute budget
- subject: POISK: consensus coverage
- subject: POISK: epitope localisation
- subject: POISK: folding protocol
- subject: POISK: scoring metric definition
- subject: POISK: contamination analysis
- subject: POISK: training data disclosure
- used in: POISK co-folding binder-decoy benchmark set (22,494 complexes)
- used in: POISK binder records with mappable patent epitope annotations (1,232)
- used in: POISK 1-in-50 screening targets (27 antigen constructs)
- part of: POISK binder-decoy separation in full-Fv ipSAE, Cohen's d
- part of: POISK binder-decoy separation in ipTM, Cohen's d
- part of: POISK pair coverage under cross-model structural consensus
- part of: POISK pairwise binder-versus-decoy accuracy under cross-model structural consensus
- part of: POISK epitope contact recovery
- part of: POISK pairwise binder-versus-decoy accuracy at 10% coverage
- part of: POISK pairwise binder-versus-decoy accuracy at 100% coverage
- part of: POISK pairwise binder-versus-decoy accuracy at 25% coverage
- part of: POISK pairwise binder-versus-decoy accuracy at 50% coverage
- part of: POISK pairwise binder-versus-decoy accuracy at 75% coverage
- part of: POISK pairs left undecidable by identical ipSAE
- part of: POISK 1-in-50 screening simulation, top-1 retrieval
- part of: POISK 1-in-50 screening simulation, top-10 retrieval