T4SEpp: A pipeline integrating protein language models to predict bacterial type IV secreted effectors

• DOI: 10.1016/j.csbj.2024.01.015
• Publication Date: 2024-01-23T07:36:04Z
• Authors: Yueming Hu; Yejun Wang; Xiaotian Hu; Haoyu Chao; Sida Li; Qinyang Ni; Yanyan Zhu; Yixue Hu; Ziyi Zhao; Ming Chen
• Keywords: T4SS; 0303 health sciences; 03 medical and health sciences; T4SE Prediction; Helicobacter pylori T4SEs; Deep learning; Protein language model; T4SEpp; TP248.13-248.65; Biotechnology; Research Article

Many pathogenic bacteria use type IV secretion systems (T4SSs) to deliver effectors (T4SEs) into the cytoplasm of eukaryotic cells, causing diseases. The identification is a crucial step in understanding mechanisms bacterial pathogenicity, but this remains major challenge. In study, we used full-length embedding features generated by six pre-trained protein language models train classifiers predicting T4SEs and compared their performance. We integrated three modules model called T4SEpp. first module searched for homologs known T4SEs, signal sequences, effector domains; second fine-tuned machine learning using data sequence feature; third best-performing models. T4SEpp outperformed other state-of-the-art (SOTA) software tools, achieving ∼0.98 accuracy at high specificity ∼0.99, based on assessment an independent validation dataset. predicted 13 from