Prediction of Protein and RNA Structures by Co-Evolution: Going Beyond Anecdotal Cases towards Large-Scale

Abstract

Structural characterization of many important proteins and protein complexes - typically preceding any detailed mechanistic exploration of their function- remains experimentally challenging. Novel statistical tools such as Direct Coupling Analysis (DCA) take advantage of the explosive growth of sequential databases and trace the co-evolution of amino acids to predict secondary and tertiary contacts for proteins [1] and RNAs [2]. These contacts can be exploited as spatial constraints in structure prediction workflows leading to excellent quality predictions [1,2,3,4]. We demonstrate for two-component signal transduction systems (TCS), a ubiquitous signal response system, how different sub-families of TCS can be identified based on genomic data [unpublished data]. Going beyond anecdotal cases of a few protein families, we have applied our methods to a systematic large-scale study of nearly 2000 PFAM protein families of homo-oligomeric proteins [unpublished data]. Also, we can apply DCA to infer mutational landscapes by capturing epistatic couplings between residues and can assess the dependence of mutational effects on the sequence context where they appear [5].