Toward quantum computational biomolecular structure prediction
| dc.contributor.author | Zaborniak, Tristan | |
| dc.contributor.supervisor | Stege, Ulrike | |
| dc.contributor.supervisor | Numanagić, Ibrahim | |
| dc.date.accessioned | 2025-09-08T21:31:28Z | |
| dc.date.available | 2025-09-08T21:31:28Z | |
| dc.date.issued | 2025 | |
| dc.degree.department | Department of Computer Science | |
| dc.degree.level | Doctor of Philosophy PhD | |
| dc.description.abstract | Biomolecules and their interactions form the material and processual basis underlying all biological phenomena, from photosynthesis to Alzheimer’s disease. Studying these systems is therefore central to the purview of all biological sciences. Computational biomolecular structure prediction (CBSP) supports this effort by leveraging computers to determine, model, and engineer biomolecular structures, properties, and processes—offering a powerful complement to laboratory-based methods. However, many core CBSP problems—such as finding minimum free energy or conformationally-stable structures given sequence information—are computationally challenging. These problems are typically NP-hard in their general form, while their corresponding decision variants are NP-complete. As a result, both formulations are resistant to efficient exact solution at large scales. Quantum computing, a developing computational paradigm leveraging quantum mechanics, offers a potential path forward, given recent evidence suggesting that certain quantum approaches may reduce resource demands for certain NP-hard problem families. Approaches include fully quantum algorithms, quantum-inspired classical heuristics, and hybrid quantum-classical frameworks, all of which may help address long-standing computational bottlenecks in CBSP. This dissertation offers a preliminary investigation of the practical potential of quantum computing for three core CBSP challenges—RNA folding, multi-body molecular docking, and protein design—that, despite their diverse applications, share structural features well suited to exploration by quantum optimization methods. Specifically, we cast each problem as a cost function network (CFN), and develop transformations of these CFNs to quadratic unconstrained binary optimization (QUBO) models in order to render them compatible with current quantum and quantum-inspired hardware. We argue that these transformations not only broaden the range of solvable CFNs across quantum platforms, but in some cases possess intrinsic features which may offer optimization advantages over native CFN formulations. Using a current‑generation superconducting flux‑qubit quantum annealer, we: (a) demonstrate its use for tuning free QUBO parameters against biomolecular structure data, and (b) benchmark solution quality and resource usage against optimized classical Monte Carlo methods, finding comparable performance. Finally, we package these methods into the Masala Quantum Computing Plugins library, an open‑source, modular CBSP platform that supports CFN construction, multiple QUBO encodings (one‑hot, domain‑wall, approximate‑binary, hybrid), and execution on both classical and quantum backends. Our contribution lays the groundwork for extensible, state‑of‑the‑art, quantum-compatible CBSP workflows. | |
| dc.description.scholarlevel | Graduate | |
| dc.identifier.uri | https://hdl.handle.net/1828/22736 | |
| dc.language | English | eng |
| dc.language.iso | en | |
| dc.rights | Available to the World Wide Web | |
| dc.subject | Quantum Computing | |
| dc.subject | Combinatorial Optimization | |
| dc.subject | Computational Biomolecular Structure Prediction | |
| dc.subject | RNA Folding | |
| dc.subject | Molecular Docking | |
| dc.subject | Protein Design | |
| dc.subject | Cost Function Network | |
| dc.subject | Quadratic Unconstrained Binary Optimization | |
| dc.subject | Simulated Annealing | |
| dc.subject | Quantum Annealing | |
| dc.subject | Approximate-Binary Encoding | |
| dc.title | Toward quantum computational biomolecular structure prediction | |
| dc.type | Thesis |