NSF Award
1564955
Proteins play fundamental roles in all biological processes. Complete description of protein structures and functions is a fundamental step towards understanding biological life and has various applications. Millions of protein sequences are available, but a majority of them have no experimentally-solved structures and functions. This project aims to greatly improve RaptorX, a fully-automated web server for computational prediction of protein structure and function, with the goal to deliver a long-term sustainable web portal to facilitate transformative research in biology. This web portal shall benefit a broad range of biological and biomedical applications, such as genome annotation, understanding of disease processes, drug design, precision medicine and even biomaterial and bio-energy development. The results will be disseminated to the broader community through a variety of venues: web servers, standalone software, publications and talks. Since late in 2011, RaptorX has served >25,000 worldwide users including middle- and high-school students. The standalone programs have been downloaded by >1500 worldwide users. After this project is fulfilled, RaptorX will contribute much more to the broader community. Students involved in this project will receive training in the intersection of computer science, molecular biology, biophysics, and biochemistry. Undergraduate and underrepresented students will be recruited through summer intern programs and collaborators. The research results will be integrated into course materials and used in the Illinois online bioinformatics program. <br/><br/>The RaptorX web server was originally developed for only template-based protein modeling. This project will transform RaptorX by first developing a few novel and powerful deep learning (e.g., Deep Conditional Convolutional Neural Fields) and structure learning (e.g., group graphical lasso) methods to significantly improve the accuracy of protein structure and functional prediction and then conducting an efficient implementation. The resultant RaptorX will be able to perform much more accurate prediction of protein secondary and tertiary structure, solvent accessibility and disordered regions, and the quality of a theoretical protein 3D model (in the absence of natives). This project will also expand the RaptorX server to perform contact prediction and contact-assisted protein folding for proteins without good templates. The RaptorX web server is available at http://raptorx.uchicago.edu, from which users can also download the standalone programs.
