Protein 3D Structure Identification by AlphaFold: a Physics-Based Prediction or Recognition Using Huge Databases?

Abstract

The great success of AlphaFold programs poses the questions: (i) What is the main reason for this success? (ii) What AlphaFolds does: physics-based prediction of the spatial structure of a protein from its amino acid sequence or recognition of this structure from similarity of the target sequence to some parts of sequences with already known spatial structures? The answers given here are: (i) the main reason for the AlphaFold’s success is the usage of huge databases which already cover virtually all protein superfamilies existing in Nature; (ii) using these databases, multiple sequence alignments, and coevolutionary information – like correlations of amino acid residues of the contacting chain regions – AlphaFold recognizes a spatial structure by similarity of the target sequence (or its parts) to related sequence(s) with already known spatial structures. We emphasize that this does not diminish the merit and utility of AlphaFold but only explains the basis of its success.