The most recent outbreak of coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has become a global pandemic. It has spread over more than 200 countries and caused numerous deaths worldwide. The central activities of SARS-CoV-2, including human cell invasion and viral duplication and infection, are conducted through the proteins coded by the viral genome as well as the protein-protein interactions between the virus and its human hosts. Determination of the structures, functions and interactions of protein molecules associated with coronaviruses can thus provide critically important knowledge to help elucidate and end the pandemic. This project will extend state-of-the-art structural bioinformatics methods to generate genome-wide protein structure and function models for SARS-CoV-2 and other human coronaviruses, which will help in understanding the general mechanisms and principles governing the virulence, diversity and evolution of these coronaviruses and facilitate the development of new treatments to cure infected individuals and terminate the COVID-19 pandemic. Multiple graduate and undergraduate students, including women and minorities, will be trained through participation in different Objectives of the project. The project results will be integrated with the bioinformatics core courses in the Bioinformatics and Biochemistry PhD Programs and the Museum of Natural History at the University of Michigan, with the purpose of enhancing the outreach and broad impacts of this research on both student and public education.
Accurately modeling protein structure and function has been a long-term challenge in structural bioinformatics and computational biology. A classical approach to this problem is comparative modeling, i.e., deducing information of unknown target proteins from known homologous proteins that are evolutionarily related to the targets. This approach is built on the assumption that similar sequences have similar structures and functions. Although they work well in many applications, the comparative approaches cannot be applied to effectively model proteins associated with SARS-CoV-2 and other human coronaviruses, because viral genomes are highly mutable, and many of the genes and gene products belonging to these viruses do not have close homologous templates with other species. To address these issues, this project plans to extend multiple algorithms developed in the PI?s lab, which have been designed primarily for non-homology-based protein structure and function prediction. In particular, the methods will utilize cutting-edge deep convolutional neural-network (DCNN) models to generate amino acid-level contact and distance maps in order to improve protein structure and interaction network modeling accuracy. Since the DCNN models are trained only on sequence databases, the performance of the approaches does not rely on the availability of structural and functional templates and can therefore be effectively used to model the coronavirus proteins that lack homologous templates; successfully developing these methods will also benefit the field of structural bioinformatics in general due to the importance of non-homologous protein structure and function prediction. In summary, the success of this project will result in the development of an urgently needed knowledge base to improve the understanding of fundamental principles associated with human coronaviruses and facilitate the development of new treatments for the COVID-19 pandemic. The data and methods produced by the project will be accessible to the community at https://zhanglab.ccmb.med.umich.edu/COVID-19/. This RAPID award is made by the Division of Biological Infrastructure, using funds from the Coronavirus Aid, Relief, and Economic Security (CARES) Act.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
