Computational methods have proven revolutionary for the rational design of artificial proteins and synthetic heteropolymers. However, while protein design benefits from large machine learning models trained on hundreds of thousands of experimentally-determined structures, synthetic heteropolymer design lacks sufficient training data and must instead rely on physics-based models and combinatorial optimization. When cast in such a form, these design problems are NP-hard, and with highly rugged energy landscapes. As a result, problems of realistic size are typically approached via Monte Carlo methods. Quantum computing—a developing computational paradigm grounded in quantum mechanics—offers a promising alternative for optimizing and sampling from such landscapes. Recent theoretical and experimental results suggest that quantum optimization approaches can substantially accelerate both optimization and sampling, with the added benefit of intrinsically uncorrelated samples arising from quantum randomness.
     In this talk, I will summarize our work encoding the rotamer packing and multi-body molecular docking problems in forms amenable to a variety of quantum optimization and sampling routines, and will present our results from a large-scale benchmarking study demonstrating that a current-generation quantum annealer already achieves performance comparable to Rosetta’s adaptive simulated annealing algorithm in finding optimal and near-optimal rotamer packing solutions. Additionally, I will discuss theoretical and experimental evidence that these encodings exhibit geometric properties that also benefit classical Monte Carlo methods. Finally, I will describe our development of the Masala quantum computing plugins library for rapid quantum methods prototyping and benchmarking, and outline our emerging work on two related fronts: (1) qubit- and gate-efficient encodings of graph partitioning problems, and (2) rapid sampling from physiologically relevant Boltzmann distributions via diabatic quantum annealing.

Refreshments will be served preceding the talk, starting at 2:45.