Non-coding RNA (ncRNA) plays an important role in regulating biological activities of animals and plants, and the representative ones are microRNA (miRNA) and long non-coding RNA (lncRNA). Recent research has found that predicting the interaction between miRNA and lncRNA is the primary task for elucidating their functional mechanisms. Due to the small scale of data, a large amount of noise, and the limitations of human factors, the prediction accuracy and reliability of traditional feature-based classification methods are often affected. Besides, the structure of plant ncRNA is complex. This paper proposes an ensemble deep-learning model based on convolutional neural network (CNN) and independently recurrent neural network (IndRNN) for predicting the interaction between miRNA and lncRNA of plants, namely, CIRNN. The model uses CNN to explore the functional features of gene sequences automatically, leverages IndRNN to obtain the representation of sequence features, and learns the dependencies among sequences; thus, it overcomes the inaccuracy caused by human factors in traditional feature engineering. The experiment results show that the proposed model is superior to shallow machine-learning and existing deep-learning models when dealing with large-scale data, especially for the long sequence.

非编码RNA（ncRNA）在调节动植物的生物活性中起着重要作用，代表性的是microRNA（miRNA）和长的非编码RNA（lncRNA）。最近的研究发现，预测miRNA和lncRNA之间的相互作用是阐明其功能机制的主要任务。由于数据规模小，噪声大以及人为因素的局限性，传统基于特征的分类方法的预测准确性和可靠性经常受到影响。此外，植物ncRNA的结构复杂。本文提出了一种基于卷积神经网络（CNN）和独立递归神经网络（IndRNN）的集成深度学习模型，用于预测植物的miRNA和lncRNA之间的相互作用，即CIRNN。该模型使用CNN自动探索基因序列的功能特征，利用IndRNN获取序列特征的表示形式，并了解序列之间的依赖性。从而克服了传统特征工程中人为因素造成的误差。实验结果表明，该模型在处理大规模数据（特别是长序列）时，优于浅层机器学习模型和现有的深度学习模型。