Peptriever: A Bi-Encoder approach for large-scale protein-peptide binding search

Abstract

MotivationIn peptide therapeutics, the successful interaction between a designed peptide and a specific receptor is crucial, while minimizing interactions with other receptors is equally essential. Current computational methods excel at estimating the probability of the former but estimating the latter requires excessive computational resources, making it challenging. ResultsIn this study, we propose transformers-based protein embeddings that can quickly identify and rank millions of interacting proteins. Furthermore, the proposed approach outperforms existing sequence- and structure-based methods, with a mean AUC-ROC and AUC-PR of 0.73. AvailabilityTraining data, scripts, and fine-tuned parameters are available at https://github.com/RoniGurvich/Peptriever. A live demonstration of the application can be found at https://peptriever.app/. Contacttomermrsn@gmail.com and zia.rehman@helsinki.fi