Summary
Distograms from protein structure prediction neural networks can be used as loss functions for protein design (1,2). This has been experimentally validated for both miniprotein binders and antibodies in the case of ESMFold2.
1.
Cho Y, Pacesa M, Zhang Z, Correia BE, Ovchinnikov S. Boltzdesign1: Inverting All-Atom Structure Prediction Model for Generalized Biomolecular Binder Design. openRxiv; 2025. Available from: https://doi.org/10.1101/2025.04.06.647261
2.
Candido S, Hayes T, Derry A, Rao R, Lin Z, Verkuil R, et al. Language Modeling Materializes a World Model of Protein Biology. 2026 Jun; Available from: http://dx.doi.org/10.64898/2026.06.03.729735