The invention discloses a
metabolite-
disease association prediction method, and belongs to the technical field of
bioinformatics. According to the method, a
metabolite-
disease-
gene multi-source fusion
heterogeneous network is constructed by using
metabolite omics information, not only is the relationship among different
biological substances considered, but also the problem of matrix sparsity is solved, and a local-global
feature extraction model LGDFE is proposed, so that the method has the advantages of being high in robustness and high in robustness. The deep relationship in the multi-source fusion
heterogeneous network is extracted from global and local angles, the problem that the deep relationship between metabolites and diseases is difficult to fully excavate in a traditional method is effectively solved, the
deep learning algorithm CNN is used for predicting data, features of different levels can be automatically fused, the parameter complexity of a common classifier is reduced, and the classification efficiency is improved. The unknown correlation between the metabolite and the
disease can be predicted, the prediction efficiency is improved, the correlation between the metabolite and the disease does not need to be obtained through a large number of experiments, and the experiment cost is remarkably reduced.