Hybrid quantum-classical computing for multimodal biomarker discovery

We are working on a new application of quantum technology to biomedicine, together with a highly multi-disciplinary team of collaborators. A major challenge in biomedical cancer research is to exploit multimodal data that capture the genomic and phenotypic changes in tumors to identify biomarkers of therapeutically targetable biological processes and patient outcomes. A typical multimodal cancer dataset contains relatively few patients but many features, including DNA mutations, gene expression, and pathology images. Supported by the Wellcome Leap Quantum for Biology (Q4Bio) program, we developed, refined, and experimentally instantiated a hybrid quantum-classical pipeline for multimodal cancer biomarker discovery, which we frame as a combinatorially difficult, yet simple and interpretable, feature selection problem. Across the three phases of Q4Bio (Fall 2023 – Spring 2026), we established a novel optimization formulation that leverages higher-order correlations among features, built resource-efficient quantum optimization methods that work together with classical solvers to find better solutions to these NP-hard problems, improved classical baselines, demonstrated our pipeline’s capabilities in important, real-world settings, and executed early end-to-end hardware demonstrations, while strengthening fault-tolerant resource projections. Collectively, these results position our hybrid feature selection framework as both a credible pathway toward clinically meaningful biomarker discovery and a realistic route to empirical quantum advantage via domain-specific quantum acceleration. As we move forward with support from IBM, we are advancing different aspects of the hybrid quantum-classical pipeline to effectively exploit quantum computational resources, including potentially using quantum machine learning steps in the pipeline. A key strength of our team, which includes Fred Chong (Infleqtion, UChicago Computer Science), Teague Tomesh (Infleqtion), Alex Pearson (UChicago Medicine), and Aram Harrow (MIT Physics), is the ability to co-design classical and quantum algorithms throughout the entire data pipeline, with a focus on solving biologically significant problems with IBM’s newest quantum machines