Scoring 100 million designs for one target

Abstract

To choose which designs to make, we score 100 million candidates for one target in one day on a GPU cluster. We combine a structure-prediction check with a learned binding score. Designs chosen this way bound their target three times as often as designs chosen by structure prediction alone.

Citation

Okafor L, Patel S. Scoring 100 million designs for one target. Applied Protein Design. 2025;2(1):14-27. Sample citation.