The Prediction Of Plant LncRNA And MiRNA Interaction Based On Multi Information

Plant lncRNA-miRNA interactions have important regulatory roles in plant growth and development,flowering and fruiting,and stress resistance,and are expected to be used for crop screening,genetics,and breeding.At present,research on non-coding RNAs is still largely focused on animals and humans,while research on plant RNA interactions is still in its infancy,and the mechanisms of these interactions are still unclear.Due to the low sequence conservation of lncRNAs,the findings in animals are not applicable to plants,and therefore the interaction mechanisms in plants need to be explored independently.Only a few methods have been proposed for plant lncRNA-miRNA interactions,and there is still a wide scope for exploration in this field.In this paper,we propose two types of methods to predict plant lncRNA-miRNA interactions based on combinatorial information,which are mainly as follows.(1)Linear neighborhood propagation model based on combinatorial information.Existing methods are basically based on machine learning and deep learning,which have the deficiencies of weak interpretation and dependence on specific features,so this paper proposes MILNP,a model constructed based on similarity networks,to construct combinatorial similarity using the common features of lncRNA and miRNA of plants and combinatorial information of interactions,and obtain prediction results by label propagation.Among them,the proposed similarity measurement algorithm ILNS employs linear reconstruction of data points in space with neighborhoods adjacent to both cosine and Euclidean distances.The results show that MILNP can predict the possible interactions between lncRNAs and miRNAs in plants with better performance than existing methods,reflecting the advantages of combined information and the accuracy of the proposed similarity measurement algorithm.(2)Integrated deep learning model based on combinatorial information.Deep learning is the mainstream of current research,and this paper uses a convolutional neural network module CNN and a stacked noise reduction autoencoder module SDAE to build an integrated model PSC.Convolutional neural networks are powerful but may lose information during sampling,while autoencoder has the advantage of preserving information,and the two can complement each other.PSC uses the combined information of plant lncRNA and miRNA sequences and sequence features,which are processed and fed into the CNN module and SDAE module,respectively,and the decisions of both modules are confused to form the final prediction.The results show that PSC excels in deep learning methods,not only with better performance,but also with simple and fast network structure,reflecting the advantages of combined information and integrated models.