PHAT: interpretable prediction of peptide secondary structures

19 Peptides have recently emerged as therapeutic molecules against various diseases, and the 20 secondary structure of peptides is a crucial determinant of their bioactivity. However, 21 accurately predicting peptide secondary structures remains a challenging task due to the lack 22 of peptide sequence data and low prediction efficiency caused by limitations in feature 23 engineering. Therefore, we developed PHAT, a deep learning framework based on a 24 hypergraph multi-head attention network and transfer learning for the prediction of peptide 25 secondary structures. Comparative results demonstrated the outstanding performance and 26 robustness of PHAT. In particular, PHAT automatically learns a set of biologically meaningful 27 knowledge on secondary sub-structures, overcoming the limitations of “black-box” in deep 28 learning-based models and providing good interpretability. Additionally, we demonstrated that 29 the structure information derived from PHAT significantly improved the performance of 30 downstream tasks such as the prediction of peptide toxicity and protein-peptide binding sites. 31 Importantly, we further explored the applicability of PHAT for contact map prediction, which 32 can aid in the reconstruction of peptide 3-D structures, thus highlighting the versatility of our 33 model. To facilitate the use of PHAT, we establish an online server which is accessible via 34 https://server.wei-group.net/PHAT/. We expect our work to assist in the design of functional 35 peptides and contribute to the advancement of structural biology research. 36 37