PmliHFM: Predicting Plant miRNA-lncRNA Interactions with Hybrid Feature Mining Network

• Published in: Interdisciplinary sciences : computational life sciences 2023-03, Vol.15 (1), p.44-54
• Main Authors: Chen, Lin, Sun, Zhan-Li
• Format: Article
• Language: English
• Subjects: Biological activity; Biomedical and Life Sciences; Coding; Computational Biology - methods; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Appl. in Life Sciences; Computer applications; Datasets; Feature extraction; Health Sciences; Information processing; Life Sciences; Mathematical and Computational Physics; Medicine; MicroRNAs; MicroRNAs - genetics; miRNA; Modules; Non-coding RNA; Original Research Article; RNA, Long Noncoding - genetics; Statistics for Life Sciences; Theoretical; Theoretical and Computational Chemistry
• ISSN: 1913-2751; 1867-1462
• DOI: 10.1007/s12539-022-00540-0

Due to the crucial role of interactions between microRNAs (miRNAs) and long non-coding RNAs (lncRNAs) in biological processes, the study of their biological functions is necessary. So far, the various computational methods have been employed to make predictions of the miRNA-lncRNA interaction, which compensate for the inadequacy of biological experiments. However, the existing methods do not consider the differences between miRNA and lncRNA in feature extraction. In this paper, we propose a hybrid feature mining network, named PmliHFM, for predicting plant miRNA-lncRNA interactions. Firstly, miRNA and lncRNA with different sequence lengths are encoded by different encodings, which can reduce the loss of information caused by using the same coding approach. Then, a hybrid feature mining network is designed to adapt to different encoding methods and extract more useful feature information than a single network. Finally, an ensemble module is utilized to integrate the training results of the hybrid feature mining network, while a prediction module is employed to determine whether there are interactions. By testing on multiple test sets, PmliHFM outperforms several state-of-the-art approaches. The results show that the AUC of PmliHFM achieves 0.8 % , 3.1 % and 0.4 % improvement respectively on three balanced datasets, and achieves 2.1 % and 1.8 % improvement respectively on two imbalanced datasets. These experiments demonstrate the feasibility of the proposed method. Graphical abstract