Evolutionary landscapes derived from amino acid sequence space serve as predictive models of functional phenotypes in protein design

Abstract

Exploring the extant sequence space of biomolecules has been useful to study molecular evolution and common statistical features between members of related families of proteins and RNA. When these statistical signatures are incorporated in the parameters of a global inference model of sequences, then multiple applications arise. These applications include the prediction of amino acid contacts in 3D space, the accurate inference of structural complexes as well as models for specificity in molecular recognition for signaling proteins and protein-RNA interactions.