Biophysics of Drug–DNA Interactions: Single-Molecule AFM, Modeling, and Machine Learning

Understanding how chemotherapeutic agents alter DNA mechanics is essential for clarifying their mechanisms of action, therapeutic efficacy, and potential side effects. We employ atomic force microscopy (AFM) to measure single-molecule force–extension curves of DNA under physiological conditions, exposed to two major classes of anticancer drugs: intercalators and cross-linkers. These high-resolution biophysical data are analyzed using worm-like chain (WLC) polymer modeling in combination with supervised machine learning (ML) to automatically classify interaction types, infer dose dependence, and extract biophysical parameters from noisy force signatures. Compared to traditional curve-fitting methods, the ML pipeline significantly improves classification accuracy, robustness, and generalizability across experimental replicates. We perform bootstrapped confidence interval analyses, ablation studies, and cross-validation to verify key predictive features and model performance. This integrated AFM–ML framework enables scalable, high-throughput characterization of drug–DNA mechanical interactions and provides a reproducible toolset for next-generation screening of DNA-targeting therapeutics.