Item

Abstract

Protein–protein interactions (PPIs) play a crucial role in numerous molecular pro-cesses. Despite many efforts, mechanisms governing molecular recognition betweeninteracting proteins remain poorly understood and it is particularly challenging to pre-dict from sequence whether two proteins can interact. Here we present a newmethod to tackle this challenge using intrinsically disordered regions (IDRs). IDRs areprotein segments that are functional despite lacking a single invariant three-dimensional structure. The prevalence of IDRs in eukaryotic proteins suggests thatIDRs are critical for interactions. To test this hypothesis, we predicted PPIs using IDRsequences in candidate proteins in humans. Moreover, we divide the PPI predictionproblem into two specific subproblems and adapt appropriate training and test strat-egies based on problem type. Our findings underline the importance of definingclearly the problem type and show that sequences encoding IDRs can aid in predict-ing specific features of the protein interaction network of intrinsically disordered pro-teins. Our findings further suggest that accounting for IDRs in future analyses shouldaccelerate efforts to elucidate the eukaryotic PPI network.