Resdiue-residue contact driven protein structure prediction using optimization and machine learning
Abstract
Significant improvements in the prediction of protein residue-residue contacts are observed in the recent years. These contacts, predicted using a variety of coevolution-based and machine learning methods, are the key contributors to the recent progress in ab initio protein structure prediction, as demonstrated in the recent CASP experiments. Continuing the development of new methods to reliably predict contact maps, tools to assess the utility of predicted contacts, and methods to construct protein tertiary structures from predicted contacts, are essential to further improve ab initio structure prediction. In this dissertation, three contributions are described -- (a) DNCON2, a two-level convolutional neural network-based method for protein contact prediction, (b) ConEVA, a toolkit for contact assessment and evaluation, and (c) CONFOLD, a method of building protein 3D structures from predicted contacts and secondary structures. Additional related contributions on protein contact prediction and structure reconstruction are also described. DNCON2 and CONFOLD demonstrate state-of-the-art performance on contact prediction and structure reconstruction from scratch. All three protein structure methods are available as software or web server which are freely available to the scientific community.
Degree
Ph. D.
Thesis Department
Rights
OpenAccess.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.