Evaluating generalization in protein–ligand cofolding methods
Peter Škrinjar, Jérôme Eberhardt, Gabriel Studer, Gerardo Tauriello, Torsten Schwede, Janani Durairaj · 2026
Read Evaluating generalization in protein–ligand cofolding methods by Peter Škrinjar, Jérôme Eberhardt, Gabriel Studer, Gerardo Tauriello, Torsten Schwede, Janani Durairaj (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13661253.txt.
Description
Skrinjar and co-authors judge in 2026 how well AI methods that fold proteins together with drug like molecules generalize to new cases. Such cofolding could speed drug discovery, but past benchmarks leaked training likeness. They built a fresh benchmark of 2,600 high grade protein ligand systems released after the methods' training cutoff.
Four leading all atom cofolding methods were tested on that set. Current methods largely memorized ligand poses from training data in their verdict. That habit limits use for truly new drug design. The benchmark gives the field a more honest yardstick, so progress claims can be checked against data the models have never seen.
Catalog record
| CREATOR | Peter Škrinjar, Jérôme Eberhardt, Gabriel Studer, Gerardo Tauriello, Torsten Schwede, Janani Durairaj |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |