Harnessing Quantum and Liquid Neural Networks for Drug Repurposing in Neurology
Don Roosan, Tiffany Khou, Saif Nirzhor, Rubayat Khan, Fahmida Hai, Inyene E. Essien-Aleksi, Andrius Baskys
Abstract
Drug development for neurological conditions faces high costs, long timelines, and frequent failures. Drug repurposing offers a solution, enhanced by advanced computational methods. Using the Broad Institute’s Drug Repurposing Hub data, we applied Quantum Neural Networks (QNNs) and Liquid Neural Networks (LNNs) to predict repurposing opportunities. Focusing on drugs’ mechanisms, clinical phases, and targets, we adapted data for quantum hardware using IBM’s Qiskit. QNNs achieved strong results (accuracy = 0.85, F1-score = 0.81, AUC-ROC = 0.88), though LNNs performed better (accuracy = 0.89, F1-score = 0.86, AUC-ROC = 0.91). Despite quantum hardware limits, QNNs show promise for drug repurposing, with potential to surpass classical models as technology advances. This study underscores the value of QNNs and LNNs in accelerating neurological drug discovery through hybrid quantum-classical approaches.