Integrating Pharmacokinetics and Pharmacodynamics Modeling with Quantum Regression for Predicting Herbal Compound Toxicity
Document Type
Article
Publication Title
arXiv
Publication Date
6-25-2025
Abstract/ Summary
Herbal compounds present complex toxicity profiles that are often influenced by both intrinsic chemical properties and pharmacokinetics (PK) governing absorption and clearance. In this study, we develop a quantum regression model to predict acute toxicity severity (LD50) for herbal-derived compounds by integrating toxicity data from NICEATM with pharmacological features from TCMSP. We first extract molecular descriptors (e.g., logP, polar surface area) alongside PK metrics such as oral bioavailability, combining them into a unified feature set. A quantum linear systems algorithm is then applied to solve the regression problem in a high-dimensional quantum state space, capturing multifaceted feature interactions efficiently. Comparative evaluation against classical models, including linear regression and random forest, shows that the quantum model achieves lower prediction errors and higher explanatory power. Analysis of learned coefficients reveals the importance of PK features for modeling toxicity, highlighting that well-absorbed, lipophilic compounds display heightened risk. We further demonstrate the model’s utility by predicting toxicity for additional herbal compounds lacking experimental data, identifying several high-risk candidates. This work underscores the potential of integrating pharmacokinetics into quantum machine learning to elucidate toxicity mechanisms, offering a more comprehensive approach to herbal compound safety assessment.