Скачать или смотреть "Designing High Affinity Drugs with AI"

NotebookLM: " ... [Presented is] a comprehensive view of advanced computational methods in biomolecular design and prediction, largely leveraging deep learning and generative models. Several papers introduce new frameworks and models, such as AtomWorks and RosettaFold-3 (RF3) for accelerating general biomolecular modeling, BoltzGen for universal binder design (proteins, peptides, nanobodies), and OMTRA for multi-task structure-based drug design, including docking and de novo design. Other sources focus on specific prediction tasks, including RMSF-net for predicting protein flexibility using cryo-EM data and PLM-interact for improving protein-protein interaction predictions, even across different species. A further theme involves creating high-quality datasets and rigorous evaluation for computational biology, exemplified by the Predicted and Experimental Peptide Binding Information (PEPBI) database and the GEMS model, which was validated on a filtered dataset to ensure genuine generalization capability in binding affinity prediction."

https://pubmed.ncbi.nlm.nih.gov/41000976/
https://doi.org/10.1101/2025.11.20.689494
https://www.nature.com/articles/s41597-025...