Thumbnails:
List:
Year:
Category:
Session:
Poster:
Getting poster data...
Javier S. Utgés, Stuart A. MacGowan, Callum M. Ives, Geoffrey J. Barton (Division of Computational Biology, School of Life Sciences, University of Dundee, Dundee, UK)
In this work we gather protein-ligand interactions data from 35 different proteins and examine the distribution of human population variants across ligand binding sites. Population variants are not randomly distributed along the genome but are constrained by protein structure and function. We propose a new approach that uses protein-ligand interactions to define binding sites on the protein. The defined sites, as well as the residues within, are characterised in terms of amino acid conservation, variation, and surface accessibility. The most conserved sites across homologues and depleted in missense variation in human are known to be functional in the target proteins. Furthermore, we group sites according to their solvent accessibility, which results in clusters of ligand sites with different patterns of site size, conservation, variation, exposure, as well as enrichment in function. Machine learning models are presented that predict these cluster labels and represent the likelihood of ligand sites being biologically relevant. We aim to shed light on those sites that are unknown to be functional, as well as to highlight positions with a particular variation profile.