Data Study Group Network Final Report: Woolfson Laboratory

Machine learning for protein folding

This report presents the outputs of a week-long collaboration between The Alan Turing Institute and Woolfson laboratory in the School of Chemistry at the University of Bristol, to predict different states of a type of protein fold called coiled coils (CCs) (Woolfson (2017) which are the structural motifs that consist of two or more α-helices winding around each other, from linear sequences of amino acids by machine learning methods.

The main objective of the data study group was to develop a machine learning model that can, if given a sequence of amino acids, be able to predict whether this sequence will participate in either a) a parallel dimer, b) an anti-parallel dimer or c) any other structure such as a trimer. It was neglected whether this sequence folds with another homomer or heteromer, or whether it binds with a sequence from the same or a different protein chain. Also the orientation for any non-dimer structure was neglected.

The second objective was to predict the above-mentioned classes using the sequence and the information about whether a sequence binds with another homomer or a heteromer.