Hermes: Large DEL Datasets Train Generalizable Protein-Ligand Binding Prediction Models
Maxwell Kleinsasser, Brayden J. Halverson, Edward Kraft, Sean Francis-Lyon, Sarah E. Hugo, Mackenzie R. Roman, Ben Miller, Andrew D. Blevins, Ian K. Quigley · 2026
Read Hermes: Large DEL Datasets Train Generalizable Protein-Ligand Binding Prediction Models by Maxwell Kleinsasser, Brayden J. Halverson, Edward Kraft, Sean Francis-Lyon, Sarah E. Hugo, Mackenzie R. Roman, Ben Miller, Andrew D. Blevins, Ian K. Quigley (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/arxiv-2602-13503.txt.
Description
Protein-ligand binding prediction is bottlenecked by messy training data: public affinity sets mix thousands of labs and assay formats, biasing models and muddying evaluation. DNA-encoded chemical libraries offer a cleaner source, with unified protocols generating massive binding data across diverse chemistry and protein targets. Hermes is a lightweight transformer trained only on DEL screens against hundreds of protein targets, among the largest and most protein-diverse DEL training sets used for interaction modeling.
Though never trained on traditional affinity measurements, Hermes generalizes to held-out targets, new chemical scaffolds, and external benchmarks from public binding data and high-throughput screens. DEL data alone thus captures transferable interaction patterns, and the lean architecture runs fast enough for large-scale virtual screening. Maxwell Kleinsasser leads the work with eight co-authors.
Catalog record
| CREATOR | Maxwell Kleinsasser, Brayden J. Halverson, Edward Kraft, Sean Francis-Lyon, Sarah E. Hugo, Mackenzie R. Roman, Ben Miller, Andrew D. Blevins, Ian K. Quigley |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-13503