Computational prediction of susceptibility to limited proteolysis for proteins with known 3D structure

Abstract

Identification of the protease substrates is important for elucidating mechanisms of many molecular processes in the living cell, including apoptosis, cell proliferation, protein activation or degradation. Computational predictions of proteolytic events can substantially reduce the amount of experimental work required for the identification of protease substrates. Although, the availability of the three‐dimensional structures for potential protease substrates facilitates cleavage sites predictions, most of the existing bioinformatics methods use the protease sequence specificity to predict proteolytic sites and rarely use structural information. To our knowledge, there is only one method, which use 3D structures of potential protease substrates, and there are no methods for characterizing the susceptibility of protein regions to proteolytic processing. To fill the gap in this filed, we developed a bioinformatic algorithm for estimating susceptibility of proteins to limited proteolysis based on known 3D structure. We used knowledge from our early study, which has been elucidated the determinants of limited proteolysis in different types of secondary structure using structural information from the proteolytic event database. We created separated machine learning models for each type of the secondary structure and combined them into a single prediction model, which is able to estimate susceptibility of peptide bonds to proteolysis for any protein with known three‐dimensional structure.