DeepAFM: Deep Learning Model Developed to Estimate Protein Structures from HS-AFM Images

Researchers at Tokyo University of Science and partners developed 'DeepAFM,' a deep learning model combining molecular dynamics simulations and AI to analyze High-speed atomic force microscopy images. By training on synthetic data, DeepAFM effectively removes noise and identifies protein conformational states, such as open or closed movements. This tool addresses the low-resolution challenges of HS-AFM, significantly advancing the real-time observation of biological molecular functions.