Using machine-learning-driven approaches to boost hot-spot's knowledge

Abstract

Understanding protein–protein interactions (PPIs) is fundamental to describe and to characterize the formation of biomolecular assemblies, and to establish the energetic principles underlying biological networks. One key aspect of these interfaces is the existence and prevalence of hot-spots (HS) residues that, upon mutation to alanine, negatively impact the formation of such protein–protein complexes. HS have been widely considered in research, both in case studies and in a few large-scale predictive approaches. This review aims to present the current knowledge on PPIs, providing a detailed understanding of the microspecifications of the residues involved in those interactions and the characteristics of those defined as HS through a thorough assessment of related field-specific methodologies. We explore recent accurate artificial intelligence-based techniques, which are progressively replacing well-established classical energy-based methodologies. This article is categorized under: Data Science > Databases and Expert Systems Structure and Mechanism > Computational Biochemistry and Biophysics Molecular and Statistical Mechanics > Molecular Interactions.