CryptoBench: cryptic protein–ligand binding sites dataset and benchmark
Vít Škrhák, Marian Novotný, Christos P Feidakis, Radoslav Krivák, David Hoksza · 2025
Read CryptoBench: cryptic protein–ligand binding sites dataset and benchmark by Vít Škrhák, Marian Novotný, Christos P Feidakis, Radoslav Krivák, David Hoksza (2025) 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-11725321.txt.
Description
Vít Škrhák and four colleagues introduce CryptoBench, a dataset and benchmark for cryptic protein binding sites, in a paper from twenty twenty-four. Structure-based site predictors are usually trained on ligand-bound protein shapes. That flatters their performance, especially for cryptic sites that appear only in free proteins. A fair test needs unbound structures as the starting point.
The dataset assembles over eleven hundred structures from bound-unbound protein pairs. Pairs were grouped, clustered by sequence, and filtered for real binding-site change. Predefined cross-validation splits make it the largest cryptic-site dataset to date. Baseline tests compared a top structure-based detector, a general site predictor, and a new sequence-based network. The sequence approach, using protein language model embeddings, outperformed expectations. The authors offer the benchmark to drive honest method development.
Catalog record
| CREATOR | Vít Škrhák, Marian Novotný, Christos P Feidakis, Radoslav Krivák, David Hoksza |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |