Abstract
Motivation: long noncoding RNA (lncRNA) is an important regulatory element in biological processes. LncRNAs share similar sequence characteristics with messenger RNAs (mRNAs) while play completely different roles, hence, it gave novel insights to biological studies. % through similarities and differences between protein coding transcripts. The development of next generation sequencing (NGS) helped discovering novel lncRNA transcripts. However, the experimental verification of numerous transcriptomes was time-consuming and costly. To alleviate these issues, a computational approach was needed to distinguish lncRNAs from the transcriptomes.
Results: We present a deep learning-based approach, lncRNAnet, to discover lncRNAs that incorporates recurrent neural networks (RNNs) for RNA sequence modeling and convolutional neural networks (CNNs) for detecting stop codons to obtain an open reading frame (ORF) indicator. lncRNAnet performed clearly better than other tools with the sequences of short lengths where the most lncRNAs are distributed on. In addition, lncRNAnet successfully learned features and showed 7.83%, 5.76%, 5.30%, and 3.78% improvements than the alternatives on a human test set (HT) in terms of specificity, accuracy, F1-score and AUC, respectively.
Codes
lncRNAnet is currently compatible with: theano=0.9.0, keras=2.0.2.
[code]
Links
Contact Information
If you have any questions, bug reports, or suggestions, please do not hesitate to contact us.
- Theoretical aspects: Prof. Sungroh Yoon (sryoon [at] snu.ac.kr)
- For other technical inquiries: Junghwan Baek (bjw5033 [at] snu.ac.kr)