Advancing virulence factor prediction using protein language models

Abstract

Background Bacterial infections have emerged as the second leading cause of death globally, with their virulence factors (VFs) playing a critical role. Accurate prediction of VFs serves not only to elucidate the mechanisms of bacterial pathogenicity, but also offers new avenues for treating bacterial diseases. Machine learning (ML) stands out as a powerful tool for swiftly and precisely identifying VFs. However, a persistent challenge with existing ML methods is the use of outdated embedding techniques and a lack of differentiation between VFs of Gram-positive and Gram-negative bacteria. Results In this study, we introduced pLM4VF, a predictive framework that utilized ESM protein language models to extract VF characteristics of G+ and G- bacteriaseparately, and further integrated the models using the stacking strategy. The top-performing ensemble models, constructed using ESM pLMs, for both types of bacteria collectively constituted pLM4VF. Extensive benchmarking experiments on the independent test demonstrated that pLM4VF outperformed state-of-the-art methods. Biological validations through cytotoxicity and acute toxicity assays further corroborated the reliability of pLM4VF. An online tool (http://139.9.105.117:8081/) has been developed that enables inexperienced researchers on ML to obtain VFs of various bacteria at the whole-genome scale. Conclusion We believe that pLM4VF will offer substantial support in uncovering pathogenic mechanisms, developing novel antibacterial treatments and vaccines, thereby aiding in the prevention and management of bacterial diseases.