Original title: Improved Peptide Docking with Privileged Knowledge Distillation using Deep Learning
Authors: Zicong Zhang,Jacob Verburgt,Yuki Kagaya,Charles Christoffer,Daisuke Kihara
An article discusses the importance of protein-peptide interactions in biological processes and the potential for understanding and altering these interactions to impact their biological functions. The article mentions the development of various computational methods for modeling receptor-peptide complexes, with recent advancements in deep learning methods, particularly through the use of AlphaFold (AF) and AlphaFold-Multimer (AFM). While AFM has shown competitive performance in modeling protein-peptide interactions, there is still room for improvement. The article introduces DistPepFold, a new approach that enhances protein-peptide complex docking by using an AFM-based architecture. DistPepFold utilizes a teacher model that incorporates native interaction information and transfers its knowledge to a student model through a distillation process. The article highlights how DistPepFold outperforms AFM in docking performance and demonstrates that the student model can learn from the teacher model to improve structural predictions based on AFM’s capabilities. The authors declare no competing interests.
Original article: https://www.biorxiv.org/content/10.1101/2023.12.01.569671v1