Summary
Structure prediction outputs from Structure prediction methods using protein conformational samplers tend to be less accurate (i.e., higher RMSD, lower TM-score) than those from vanilla structure prediction neural networks like AlphaFold and ESMFold (1,2). This also includes features like broken chains (what (1) calls “chain ruptures”).
Figures
Ref (1)
1.
Stratiichuk R, Kyrylenko R, Koleiev I, Savchenko I, Voitsitskyi T, Husak V, et al. Sampling and Ranking of Protein Conformations Using Machine Learning Techniques Do Not Improve the Quality of Rigid Protein–Protein Docking. Journal of Chemical Information and Modeling. 2025;65(19):10167–79. Available from: https://doi.org/10.1021/acs.jcim.5c01765
2.
Jing B, Berger B, Jaakkola T. AlphaFold Meets Flow Matching for Generating Protein Ensembles. In: GenBio@NeurIPS2023. 2023. Available from: https://openreview.net/forum?id=yQcebEgQfH