Модель PPFlow использует передовые методы машинного обучения для проектирования терапевтических пептидов с заданными свойствами. Благодаря архитектуре на базе условного сопоставления потоков (flow matching), этот ИИ эффективно моделирует геометрию молекул, открывая новые возможности для создания лекарств.
Therapeutic peptides have proven to have great pharmaceutical value and potential in recent decades. However, methods of AI-assisted peptide drug discovery are not fully explored. To fill the gap, we propose a target-aware peptide design method called PPFlow, based on conditional flow matching on torus manifolds, to model the internal geometries of torsion angles for the peptide structure design. Besides, we establish a protein-peptide binding dataset named PPBench2024 to fill the void of massive data for the task of structure-based peptide drug design and to allow the training of deep learning methods. Extensive experiments show that PPFlow reaches state-of-the-art performance in tasks of peptide drug generation and optimization in comparison with baseline models, and can be generalized to other tasks including docking and side-chain packing.