As a result of the rapid development of next generation sequencing techniques, we now have access to hundreds and often many thousands of sequences which belong to the same family. Such a large amount of sequence data for a particular protein family, along with recent developments in computational statistics, enables an entirely new kind of evolutionary analysis to be performed on sequences, where for the first time we can build statistically significant networks of correlated mutations i.e. functionally important mutations occurring at one position in the protein with mutations occurring at other positions. These techniques enable entirely new applications of sequence data such as de novo protein structure prediction for proteins of almost any size, inference of protein-protein and domain-domain interactions, prediction of protein function and even prediction of allosteric changes in protein conformation upon binding. In this talk I will be presented results of some of our recent work in this area, in particular looking at various machine learning approaches to extending the applicability and usefulness of these methods.
‘Predicting Protein Structure and Protein-Protein Interactions using Amino Acid Coevolution and Machine Learning’
Thursday, December 1, 2016 - 15:00
MSI Small Lecture Theatre
Professor Geoff Barton FRSE FRSB
Prof David T. Jones
Francis Crick Institute and UCL