This project aims to improve the accuracy and decrease the computational cost of computational methods for designing peptides and proteins for medical and manufacturing applications.
Artificial proteins and peptides represent a powerful and versatile class of molecules. The sequence of amino acid building blocks in such a molecule uniquely determines the fold of the molecule, and the fold uniquely determines the molecule’s function. Unfortunately, the sequence-fold-function relationship is one that is computationally challenging to untangle, due the vastness of both the possible sequence space and the possible conformational space.
This project aims to reduce the computational and energetic costs of producing successful peptide macrocycle drugs or industrial enzymes using two approaches. First, it will develop low-cost machine learning methods that can approximate the output of computationally expensive validation simulations, ultimately allowing users without access to large-scale resources to perform design and validation tasks on much more modest computing systems. Second, it will explore the use of quantum computing technologies as a means of solving the design problem at much lower energetic cost. This requires large-scale classical computing hardware both for carrying out quantum computing simulations during quantum algorithm development, and for performing computational validation of designs produced on current-generation quantum computing hardware.
Ultimately, this project aspires to decimate the computational and energetic cost of creating successful, computationally designed folding heteropolymers with useful functions in medicine and manufacturing, and to greatly enhance the accessibility of these computational design technologies to the scientific community.