Homologues not needed: Structure prediction from a protein language model.

Journal: Structure (London, England : 1993)
Published Date:

Abstract

Accurate protein structure predictors use clusters of homologues, which disregard sequence specific effects. In this issue of Structure, Weißenow and colleagues report a deep learning-based tool, EMBER2, that efficiently predicts the distances in a protein structure from its amino acid sequence only. This approach should enable the analysis of mutation effects.

Authors

  • Nir Ben-Tal
    Department of Biochemistry and Molecular Biology, George S. Wise Faculty of Life Sciences, Tel Aviv University, Tel Aviv 6997801, Israel. Electronic address: bental@tauex.tau.ac.il.
  • Rachel Kolodny
    Department of Computer Science, Jacobs Building, University of Haifa, 199 Aba Houshi Road, Mount Carmel, Haifa, Israel 3498838.