Rosetta Energy Analysis of AlphaFold2 models: Point Mutations and Conformational Ensembles

Abstract

AlphaFold2’s ability to accurately predict protein structures from a multiple sequence alignment (MSA) has raised many questions about the utility of the models generated in downstream structural analysis. Two outstanding questions are the prediction of the consequences of point mutations and the completeness of the landscape of protein conformational ensembles. We previously developed a method, SPEACH_AF, to obtain alternate conformations by introducing residue substitutions across the MSA and not just within the input sequence. Here, we compared the structural and energetic consequences of having the mutation(s) in the input sequence versus in the whole MSA (SPEACH_AF). Both methods yielded models different from the wild-type sequence, with more robust changes when the mutation(s) were in the whole MSA. To evaluate models of conformational diversity, we used SPEACH_AF and a new MSA subsampling method, AF_cluster, combined with model relaxation in Rosetta. We find that the energetics of the conformations generated by AlphaFold2 correspond to those seen in experimental crystal structures and explored by standard molecular dynamic methods. Combined, the results support the fact that AlphaFold2 can predict structural changes due to point mutations and has learned information about protein structural energetics that are encoded in the MSA.