PLEKv2: predicting lncRNAs and mRNAs based on intrinsic sequence features and the coding-net model

• Published in: BMC genomics Vol. 25; no. 1; pp. 756 - 10
• Main Authors: Li, Aimin, Zhou, Haotian, Xiong, Siqi, Li, Junhuai, Mallik, Saurav, Fei, Rong, Liu, Yajun, Zhou, Hongfang, Wang, Xiaofan, Hei, Xinhong, Wang, Lei
• Format: Journal Article
• Language: English
• Published: London BioMed Central 02.08.2024; BioMed Central Ltd; Springer Nature B.V; BMC
• Subjects: Accuracy; Algorithms; Animal behavior; Animal Genetics and Genomics; Animal models; Animals; Arabidopsis thaliana; Biomedical and Life Sciences; Classification; Coding-net; Comparative analysis; Computational Biology - methods; Datasets; Deep learning; Gene sequencing; Genetic research; Genomes; Human performance; Humans; Life Sciences; lncRNAs; Machine learning; Messenger RNA; Metastasis; Microarrays; Microbial Genetics and Genomics; Neural networks; Non-coding RNA; Nucleotide sequence; Nucleotides; Peptides; Performance evaluation; Plant Genetics and Genomics; PLEK; Primates; Proteins; Proteomics; RNA; RNA sequencing; RNA, Long Noncoding - genetics; RNA, Messenger - genetics; Software; Testing; China; lncRNAs; Deep learning; Coding-net; PLEK
• ISSN: 1471-2164; 1471-2164
• DOI: 10.1186/s12864-024-10662-y

Background Long non-coding RNAs (lncRNAs) are RNA transcripts of more than 200 nucleotides that do not encode canonical proteins. Their biological structure is similar to messenger RNAs (mRNAs). To distinguish between lncRNA and mRNA transcripts quickly and accurately, we upgraded the PLEK alignment-free tool to its next version, PLEKv2, and constructed models tailored for both animals and plants. Results PLEKv2 can achieve 98.7% prediction accuracy for human datasets. Compared with classical tools and deep learning-based models, this is 8.1%, 3.7%, 16.6%, 1.4%, 4.9%, and 48.9% higher than CPC2, CNCI, Wen et al.’s CNN, LncADeep, PLEK, and NcResNet, respectively. The accuracy of PLEKv2 was > 90% for cross-species prediction. PLEKv2 is more effective and robust than CPC2, CNCI, LncADeep, PLEK, and NcResNet for primate datasets (including chimpanzees, macaques, and gorillas). Moreover, PLEKv2 is not only suitable for non-human primates that are closely related to humans, but can also predict the coding ability of RNA sequences in plants such as Arabidopsis . Conclusions The experimental results illustrate that the model constructed by PLEKv2 can distinguish lncRNAs and mRNAs better than PLEK. The PLEKv2 software is freely available at https://sourceforge.net/projects/plek2/ .