Automatic Identification of Fold in Optical Tweezers‐Based Single‐Molecule Force Spectroscopy

Abstract Single‐Molecule Force Spectroscopy (SMFS) provides crucial insights into molecular mechanical properties and interactions. However, the localization of folding events relies on manual operations, which are inefficient and susceptible to subjective errors. Existing automated methods also face limitations in accuracy and robustness. To overcome these challenges, this study introduces an Inception‐based Self‐attention Skip Network (ISSN), combining Inception blocks with a Self‐attention mechanism to classify and localize folding events, including initiation and termination sites. And proposed a Force‐Distance Curve (FDC) simulation method to address the issue of insufficient datasets. Validation on Deoxyribonucleic Acid (DNA) hairpin Force‐Distance Curves (FDCs) demonstrates that ISSN achieves outstanding performance, including 99.8% classification accuracy, a low mean absolute error (MAE) of 0.176 pN for the force site, and 1.57 nm MAE for the distance site, even under high noise conditions. Compared to other classic models, ISSN not only delivers superior accuracy but also exhibits strong generalization and robustness for automated optical tweezers SMFS analysis.