LncLocFormer: a Transformer-based deep learning model for multi-label lncRNA subcellular localization prediction by using localization-specific attention mechanism

• DOI: 10.1093/bioinformatics/btad752
• Publication Date: 2023-12-18T20:46:16Z
• Authors: Min Zeng; Yifan Wu; Yiming Li; Rui Yin; Chengqian Lu; Junwen Duan; Min Li
• Keywords: Original Paper; 0303 health sciences; 03 medical and health sciences; Deep Learning; Computational Biology; RNA, Long Noncoding; Software

There is mounting evidence that the subcellular localization of lncRNAs can provide valuable insights into their biological functions. In real world transcriptomes, are usually localized in multiple localizations. Furthermore, have specific patterns for different Although several computational methods been developed to predict lncRNAs, few them designed localizations, and none take motif specificity consideration.In this study, we proposed a novel deep learning model, called LncLocFormer, which uses only lncRNA sequences multi-label localization. LncLocFormer utilizes eight Transformer blocks model long-range dependencies within sequence shares information across sequence. To exploit relationship between localizations find distinct employs localization-specific attention mechanism. The results demonstrate outperforms existing state-of-the-art predictors on hold-out test set. conducted analysis found capture known motifs. Ablation studies confirmed contribution mechanism improving prediction performance.The web server available at http://csuligroup.com:9000/LncLocFormer. source code be obtained from https://github.com/CSUBioGroup/LncLocFormer.