MicroRNAs (miRNAs) and long non-coding RNAs (lncRNAs) are both non-coding RNAs (ncRNAs) and their interactions play important roles in biological processes. Computational methods, such as machine learning and various bioinformatics tools, can predict potential miRNA–lncRNA interactions, which is significant for studying their mechanisms and biological functions. A growing number of RNA interaction predictors for animal have been reported, but they are unreliable for plant due to the differences of ncRNAs in animal and plant. It is urgent to build a reliable plant predictor, especially for cross-species. This paper proposes an ensemble deep learning model based on multi-level information enhancement and greedy fuzzy decision (PmliPEMG) for plant miRNA–lncRNA interaction prediction. The fusion complex features, multi-scale convolutional long short-term memory networks, and attention mechanism are adopted to enhance the sample information at the feature, scale, and model levels, respectively. An ensemble deep learning model is built based on a novel method (greedy fuzzy decision) which greatly improves the efficiency. The multi-level information enhancement and greedy fuzzy decision are verified to have the positive effects on prediction performance. PmliPEMG can be applied to the cross-species prediction. It shows better performance and stronger generalization ability than state-of-the-art predictors and may provide valuable references for related research.

MicroRNAs (miRNAs) 和长链非编码 RNAs (lncRNAs) 都是非编码 RNAs (ncRNAs)，它们的相互作用在生物过程中起着重要的作用。计算方法，如机器学习和各种生物信息学工具，可以预测潜在的 miRNA-lncRNA 相互作用，这对于研究它们的机制和生物学功能具有重要意义。已经报道了越来越多的动物 RNA 相互作用预测因子，但由于动植物中 ncRNA 的差异，它们对植物并不可靠。迫切需要建立一个可靠的植物预测器，特别是对于跨物种。本文提出了一种基于多级信息增强和贪婪模糊决策（PmliPEMG）的集成深度学习模型，用于植物miRNA-lncRNA相互作用预测。融合复杂特征，采用多尺度卷积长短期记忆网络和注意力机制分别在特征、尺度和模型层面增强样本信息。基于一种新的方法（贪婪模糊决策）构建集成深度学习模型，大大提高了效率。验证了多级信息增强和贪婪模糊决策对预测性能的积极影响。PmliPEMG 可应用于跨物种预测。它比最先进的预测器表现出更好的性能和更强的泛化能力，可为相关研究提供有价值的参考。基于一种新的方法（贪婪模糊决策）构建集成深度学习模型，大大提高了效率。验证了多级信息增强和贪婪模糊决策对预测性能的积极影响。PmliPEMG 可应用于跨物种预测。它比最先进的预测器表现出更好的性能和更强的泛化能力，可为相关研究提供有价值的参考。基于一种新的方法（贪婪模糊决策）构建集成深度学习模型，大大提高了效率。验证了多级信息增强和贪婪模糊决策对预测性能的积极影响。PmliPEMG 可应用于跨物种预测。它比最先进的预测器表现出更好的性能和更强的泛化能力，可为相关研究提供有价值的参考。