Speaker
Description
Motivation
Drug lead optimization requires navigating a chemical space estimated at 10⁶⁰ molecules. Classical approaches like evolutionary algorithms, Bayesian optimization, and reinforcement learning build exploration distributions from past fitness evaluations, so they cannot amplify probability toward unobserved, high-fitness regions, a structural limitation, not a scaling one.
Approach
Starting from a known active molecule, we build a local molecular subspace as a structured graph: nodes are synthetically accessible compounds from standard medicinal chemistry transformations, edges connect molecules reachable by one transformation step. Subspace size is capped by a bounded qubit and depth budget. A Quantum Walk Optimization Algorithm (QWOA) is deployed on this graph: a fitness oracle encodes molecular quality as a phase, and alternating walk and phase-separation unitaries drive amplitude concentration toward high-fitness regions through constructive interference. Circuit depth follows from the graph's structure and the layers needed for convergence. Measurement yields a distribution biased toward promising molecules, without gradient information or exhaustive enumeration.
Expected results
The subspace stays small enough to be classically precomputable and circuit-compilable, within the NISQ-feasible regime, grounded in bounded register size and depth, not qubit count alone. The parameter schedule is derived analytically from graph structure, eliminating the outer variational loop and its trainability issues, yielding a non-variational, hardware-ready sampler.
Significance
This is the first application of QWOA to chemical space exploration as combinatorial search over synthetically accessible molecules, establishing a hardware-feasible methodology for quantum-probabilistic lead optimization.
Keywords: quantum walks, chemical space exploration, drug discovery, lead optimization, combinatorial optimization
| Topic | Hybrid quantum-classical approaches for learning and optimization |
|---|---|
| Additional Topic | Theoretical foundations of quantum learning |