Microbe And Disease Association Prediction Algorithm

Microbes are a collective term for all tiny organisms.In recent years,more and more studies have proved that microbes play an important role in human health,immune defense,cancer control,and nutritional absorption.Identification of disease-related microbes will help people understanding of the pathogenesis of complex diseases and it can also promote the development of related drugs for the prevention,diagnosis and treatment of diseases.However,biological methods to identify diseases and microorganisms are not only costly but also time consuming.Therefore,the use of currently known microbial-disease correlation data to develop effective computational models to identify potential disease-related microorganisms will greatly reduce biological experimental samples,thereby reducing experimental cycles and costs.In this paper,we proposed two method to discover potential association data between microbes and diseases,based on KATZ model and bipartite network recommendation algorithm to predict microbial-disease association(KATZBNRA)and based on the linear model algorithm to predict microbialdisease association(LMMDA).Both methods integrate the microbial Gaussian similarity and disease Gaussian similarity into the algorithm model.Based on KATZ model and binary network recommendation algorithm,this method combines bipartite network recommendation algorithm and KATZ model,the recommendation score matrix is calculated using the binary network recommendation algorithm,and integrate the recommendation score matrix and two similar matrixes above.The integration matrix constructs a heterogeneous network of diseases and microorganisms,and brings the heterogeneous network into the KATZ model to realize the prediction of the microorganism-disease association;The linear model-based algorithm directly integrates known related microbial-disease data with two similarity matrices,uses the linear model to calculate the association score,and then uses the network projection algorithm to further optimize the correlation score to achieve microbial-disease correlation.Prediction.The two methods proposed in this paper use leave-one-out cross validation(LOOCV),5-fold cross validation,and 2-fold cross validation to evaluate the performance of the algorithm.At the same time,the key parameters in the algorithm are also crossvalidated,and their AUC values are obtained.The AUC values of LOOCV,five-fold cross validation,and two-fold cross validation of KATZBNRA are 0.9098,0.8972,and 0.8463,respectively,and the AUC values of LMMDA are 0.8923,0.8897,and 0.8763,respectively.The results show the prediction results of these two methods Have higher credibility.At the same time,the text also makes a case analysis of specific diseases to further evaluate the accuracy of the algorithm.