AlphaFold-predicted Protein Structure vs Experimentally Obtained Protein Structure: An Emphasis on the Side Chains

Daiki Shiono, Takashi Yoshidome

Research output: Contribution to journalArticlepeer-review

Abstract

AlphaFold is a neural-network model that enables a highly accurate prediction of the three-dimensional structure of a protein from its amino acid sequence. However, the prediction performance is very good for the backbone but is unclear for the side chains. Hence, in this study, we compared the side-chain conformations between the AlphaFold-predicted and cryo-electron microscopy structures of the β subunit of the F1-ATPase. The results showed that χ1-angle values were highly different for 93 residues among 484 residues: The differences of χ1-angle values for these residues were larger than 70°. It can be concluded from the results that the prediction performance of AlphaFold is not high for the side-chain conformations. In addition, we discussed the possibility of eliminating the difference in side-chain conformations using conventional molecular dynamics simulations and suggested that generalized-ensemble algorithms would be required for elimination.

Original languageEnglish
Article number064804
Journaljournal of the physical society of japan
Volume91
Issue number6
DOIs
Publication statusPublished - 2022 Jun 15

ASJC Scopus subject areas

  • Physics and Astronomy(all)

Fingerprint

Dive into the research topics of 'AlphaFold-predicted Protein Structure vs Experimentally Obtained Protein Structure: An Emphasis on the Side Chains'. Together they form a unique fingerprint.

Cite this