Deep learning tools are top performers in long non-coding RNA prediction

• DOI: 10.1093/bfgp/elab045
• Publication Date: 2021-12-03T20:10:40Z
• Authors: Tea Ammunét; Ning Wang; Sofia Khan; Laura L Elo
• Keywords: Robustness

The increasing amount of transcriptomic data has brought to light vast numbers potential novel RNA transcripts. Accurately distinguishing long non-coding RNAs (lncRNAs) from protein-coding messenger (mRNAs) challenged bioinformatic tool developers. Most recently, tools implementing deep learning architectures have been developed for this task, with the discovering sequence features and their interactions still not surfaced in current knowledge. We compared performance other predictive that are currently used lncRNA coding prediction. A total 15 representing variety available methods were investigated. In addition known annotated transcripts, we also evaluated use actual studies real-life data. robustness scalability tools' was tested varying sized test sets different proportions lncRNAs mRNAs. addition, ease-of-use each scored. Deep top performers most metrics labelled transcripts similarly dataset. However, proportion mRNAs affected all tools. Computational resources utilized differently between top-ranking tools, thus nature study may affect decision choosing one well-performing over another. Nonetheless, results suggest favouring broad use.