PepHemo-LLM: a protein large language models-based deep learning method to identify hemolytic peptides

PepHemo-LLM is based on protein large language models (e.g., ProteinBERT) to tokenize peptide sequences and predict whether a peptide is hemolytic. The approach can guide lead optimization to derive safe peptidic agents. Concurrently, such pLLM-based approaches have also been devised to model the susceptibility/resistance of AIDS patients with different HIV protease mutations to FDA approved drugs, termed HIVP-LLM, achieving 95-100% prediction accuracy. Both PepHemo-LLM and HIVP-LLM are implemented with the Flask framework, and also optimized for friendly use on mobile devices. PepHemo-LLM is evalauted based on comparison with our initial development of PepHemo by combining Chaos Game Representation (CGR) and supervised autoencoder (SAE) for identification of hemolytic peptides. They have achieved similarly high prediction robustness and accuracy.

Input: Peptide Sequence(s)