Quantum-Driven Predictive Analytics for Precision Medicine and Pharmacovigilance: A Multimodal Simulation Framework for Early Disease and Adverse Drug Event Detection

Résumé fourni par la source

The computational intractability of the analysis of multimodal large-dimensional data, including genomics, electronic health records (EHR), proteomics, and real-world evidence, based on classical machine learning (ML) models, is the fundamental limitation of the precision medicine and proactive pharmacovigilance paradigm. Such models can also have problems modeling complex, non-linear feature interactions important to detect diseases at an early stage and to detect rare adverse drug events (ADEs). To address these shortcomings, this paper is suggesting a new hybrid quantumclassical framework of computation. We make use of the unique ability of the quantum computation, like superposition and entanglement to create more powerful predictive models. The encoding of the multimodal clinical and pharmacological data into quantum states (via specialized feature maps) followed by processing by Variational Quantum Classifiers (VQC) and Quantum Support Vector Machines (QSVM) on simulated quantum processors, is our framework. To benchmark, we made two major tasks: Type 2 Diabetes (T2D) early-stage prediction based on synthetic multimodal data and identification of rare ADE based on a curated selection of FDA Adverse Event Reporting System (FAERS). The outcome of the simulations indicates that our quantum-based models are always more effective than the classical models, such as the XGBoost and Deep Neural Networks, especially when it comes to sensitivity and F1 score in imbalanced class issues. QSVM model obtained an AUC ROC that is 7.3 percent greater than the optimal classical model, and a 15 percent greater recall than the optimal classical model in T2D prediction and rare ADE detection, respectively. This data supports the fact that quantum machine learning (QML) can be used to bring in a new dawn of predictive analytics in healthcare, and that it provides the way to more accurate, comprehensive, and proactive medical care.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.