基于宿主—微生物网络的疾病预测算法OA

An algorithm for disease prediction based on a host-microbe dual-layer network was proposed to address the issue of microbial interactions being overlooked in disease pre-diction methods.Host-microbe networks were constructed by microbial sequencing data and phylogenetic trees.The restart random walk algorithm was applied to propagate sig-nals across the network,and disease-associated microbial motifs were identified with prior microbial knowledge.These motifs were converted into predictive features and incorporat-ed into a model for disease prediction.The experimental results show that in the detection of colorectal cancer,the AUC value of logistic regression increases from 0.84 to 0.90,and the AUC value of random forest increases from 0.88 to 0.92.