LncPNdeep: A long non-coding RNA classifier based on Large Language Model with peptide and nucleotide embedding

Abstract

Abstract Long non-coding RNA plays an important role in various gene transcription and peptide interactions. Classifying lncRNAs from coding RNA is a crucial step in bioinformatics analysis which seriously affects the post-analysis for transcriptome annotation. Although several machine learning-based methods were developed to classify lncRNAs, these methods were mainly focused on nucleotide features without considering the information from the peptide sequence. To integrate both nucleotide and peptide information in lncRNA classification, one efficient deep learning is desired. In this study, we developed one concatenated deep neural network named LncPNdeep to combine this information. LncPNdeep incorporates both peptide and nucleotide embedding from masked language modeling (MLM), being able to discover complex associations between sequence information and lncRNA classification. LncPNdeep achieves state-of-the-art performance in the human transcript database compared with other existing methods (Accuracy=97.1%). It also exhibits superior generalization ability in cross-species comparison, maintaining consistent accuracy and F1 scores compared to other methods. The combination of nucleotide and peptide information makes LncPNdeep able to facilitate the identification of novel lncRNA and gain high accuracy for classification. Our code is available at https://github.com/yatoka233/LncPNdeep