Coarse-grained lattice protein folding on a quantum annealer

TL;DR

This paper advances quantum annealing methods for lattice protein folding, reducing circuit complexity and successfully folding larger proteins like Chignolin and Trp-Cage, demonstrating improved accuracy and scalability.

Contribution

It introduces improved Ising Hamiltonian encodings and a heuristic for splitting large problems, enabling folding of larger proteins on current quantum annealing hardware.

Findings

Reduced quantum circuit complexity from quadratic to quasilinear

Successfully folded Chignolin (10 residues) on a planar lattice

Folded Trp-Cage (8 residues) on a cubic lattice

Abstract

Lattice models have been used extensively over the past thirty years to examine the principles of protein folding and design. These models can be used to determine the conformation of the lowest energy fold out of a large number of possible conformations. However, due to the size of the conformational space, new algorithms are required for folding longer proteins sequences. Preliminary work was performed by Babbush et al. (2012) to fold a small peptide on a planar lattice using a quantum annealing device. We extend this work by providing improved Ising-type Hamiltonian encodings for the problem of finding the lowest energy conformation of a lattice protein. We demonstrate a decrease in quantum circuit complexity from quadratic to quasilinear in certain cases. Additionally, we generalize to three spatial dimensions in order to obtain results with higher correlation to the actual atomistic 3D structure of the protein and outline our heuristic approach for splitting large problem instances into smaller subproblems that can be directly solved with the current D-Wave 2000Q architecture. To the best of our knowledge, this work sets a new record for lattice protein folding on a quantum annealer by folding Chignolin (10 residues) on a planar lattice and Trp-Cage (8 residues) on a cubic lattice.