LassoESM a tailored language model for enhanced lasso peptide property prediction

作者：Xuenan Mi, Susanna E. Barrett, Douglas A. Mitchell, Diwakar Shukla · 发表于：Nature Communications · 年份：2025 · DOI：10.1038/s41467-025-63412-3 · 被引用次数：4 · 研究领域：Chemical Synthesis and Analysis、RNA and protein synthesis mechanisms、vaccines and immunoinformatics approaches

Ribosomally synthesized and post-translationally modified peptides (RiPPs) are a diverse group of natural products. The lasso peptide class of RiPPs adopt a unique [1]rotaxane conformation formed by a lasso cyclase, conferring diverse bioactivities and remarkable stability. The prediction of lasso peptide properties, such as substrate compatibility with a particular lasso cyclase or desired biological activity, remains challenging due to limited experimental data and the complexity of substrate fitness landscapes. Here, we develop LassoESM, a tailored language model that improves lasso peptide property prediction. LassoESM embeddings enable accurate prediction of substrate compatibility, facilitate identification of novel non-cognate cyclase-substrate pairs, and enhance prediction of RNA polymerase inhibitory activity, a biological activity of several known lasso peptides. We anticipate that LassoESM and future iterations will be instrumental in the rational design and discovery of lasso peptides with tailored functions.