BCAMLI: A Bidirectional Cross-Attention Mechanism-Based Prediction Model for Plant miRNA-lncRNA Interaction

The prediction of plant miRNA-lncRNA interactions is vital for understanding the mechanism of miRNA-lncRNA interactions and further exploring their genomic functions. Currently, numerous computational techniques have been developed to predict plant miRNA-lncRNA interactions. However, the majority of existing approaches focus solely on miRNA and lncRNA feature extraction, ignoring the fusion learning of miRNA-lncRNA interactions, which may result in insufficient feature representation. In this paper, we present a bidirectional cross-attention mechanism-based prediction model for plant miRNA-lncRNA interactions (BCAMLI). We embed RNAs using a pre-trained model (Doc2vec), which can fully combine the context information of k-mer words with the remote dependent information of the RNA sequence, allowing us to capture sequence features with more semantic information. We develop a novel bidirectional cross-attention mechanism for enhancing interactive feature learning between miRNAs and lncRNAs, allowing us to better understand underlying biological mechanisms and improve MLI predictions. Finally, we evaluate BCAMLI on the benchmark dataset, and our proposed model outperforms the representative baseline.