Construction Of Magnaporthe Oryzae-rice Multi-omics Hierarchical Heterogeneous Interaction Network Based On Deep Learning

Recent studies have shown that fungi mainly infect plants through cross-speciesRNA interference mechanism（RNAi）and secreted protein infection mechanism.Magnaporthe oryzae sRNAs use RNAi mechanism of rice to achieve infection process,while effector proteins increase their self-toxicity by inhibiting PTI to achieve the purpose of infection.However,the specific mechanism involved in the regulation of sRNAs and proteins during the interaction between M.oryzae and rice is still poorly understood.Therefore,it is of great significance for the control of rice blast to deeply study the mechanism of M.oryzae infecting rice by combining transcriptomics and proteomics.In this paper,the genomic,transcriptomic and proteomic data during the interaction between M.oryzae and rice were analyzed in a targeted manner,and a hierarchical network of each omics was constructed horizontally.The association between multi-omics was predicted by using the deep learning method,and the hierarchical heterogeneous interaction network of M.oryzae-rice multi-omics was longitudinally constructed.The functional enrichment analysis was carried out to find the key factors involved in the M.oryzae-rice interaction mechanism.Firstly,from the genome,transcriptome and proteome data,through different analysis methods,969 rice genes,1850 M.oryzae-rice interaction pairs,660 M.oryzae internal RNA interaction pairs and 2653 protein interaction pairs between M.oryzae and rice were screened out.Then,this paper proposed two prediction models based on deep learning methods.The first is the Recurrent Neural Network-based M.oryzae-Rice Protein Interaction Prediction model（RNNMRPIP）,which predicts protein pairs that have interactions between M.oryzae and rice.The second is the Autoencoder-based M.oryzae-Rice Multi-omics Relationship Prediction model（AEMRMRP）,which predicts the multi-omics relationship between M.oryzae and rice.Finally,each omics interaction network was analyzed respectively.The M.oryzae-rice multi-omics hierarchical heterogeneous interaction network was integrated and constructed,and its GO functional modules and KEGG enrichment pathways were mined.Most of the enrichment results were found to be related to gene expression,protein synthesis and transport,and metabolic processes.Studies on M.oryzae infecting rice from the perspective of constructing a multi-omics hierarchical heterogeneous interaction network based on deep learning method have not been reported.In this paper,big data analysis and deep learning method were applied to deeply analyze the process of M.oryzae infecting rice jointly realized by sRNAs and proteins.Compared with previous single omics methods,the efficiency and accuracy of the proposed method were greatly improved.In this paper,we proposed a new approach for constructing a cross-species multi-omics interaction regulatory network and on this basis to construct a prediction model for protein interaction pairs of M.oryzae and rice and a prediction model for association relationships in multi-omics hierarchical heterogeneous interaction network.This study elucidated the key regulatory factors and mechanisms of M.oryzae infecting rice,which was helpful to formulate persistent and broad-spectrum control strategies for rice blast,and also made a certain contribution to solving the problem of other fungi invading plants.