Towards Breakthroughs in Protein Structure Calculation and Design

PI Name: 
David Baker
PI Email: 
dabaker@u.washington.edu
Institution: 
University of Washington
Allocation Program: 
INCITE
Allocation Hours at ALCF: 
120 Million
Year: 
2016
Research Domain: 
Chemistry

While progress is being made in protein structure modeling, the ability to sufficiently sample conformation space, the relative positions of all atoms in 3D, is a limiting factor in protein structure prediction and design. This project aims to advance the simulation software suite ROSETTA, a pioneering tool for computational modeling of biomolecular structures and designs. Researchers plan to optimize the ROSETTA energy function, which provides an approximation of natural protein energetics. Calibration of the energy function can be improved by using molecular simulations predicting physical observables, such as the bulk properties of liquid. Computationally intensive simulations on Mira will further optimize calibration of the ROSETTA energy function for improved accuracy and precision on smaller computing platforms.

Previously, this project described a breakthrough in conformational sampling utilizing highly parallel computations on Mira, which determined the solution structures of proteins up to 40 kilodaltons, with limited experimental data. Only large supercomputers have the capacity and rapid inter-process communication ability to consider large numbers of conformational states simultaneously, in parallel, and to permit a sufficiently fast search for an optimal sequence.

The research will investigate how to refine models from structure prediction, given information about how the protein sequence varies across related organisms. More multistate design work will focus on generating catalytic and therapeutic peptides that bind to targets of interest—including influenza and other pathogens—and a massively parallel enzyme design protocol for the de novo design of novel enzymes and catalytic sites. Using both canonical and noncanonical building blocks, ROSETTA’s parametric design tools to will help create self-assembling peptide helical bundles, small protein folds composed of several alpha helices that are usually nearly parallel or antiparallel to each other.

This research addresses contemporary issues in medicine, energy, and technology, including the development of protein therapeutics for a number of diseases, and the design of protein reagents to capture and destroy various pathogens and toxins.

Catalyst: