Research On The Distribution Characteristics Of Flora Information Based On The Probabilistic Topic Model

In the era of big data,data has become a very important production factor.Data mining has been applied to all walks of life.Among them,the excavation in the field of intestinal microbiology is the current research hotspot.Since the intestinal microflora plays an important role in the generation and treatment of human diseases,the data mining method is used to study the relationship between the information distribution characteristics of the intestinal microflora and the disease.At present,commonly used methods for studying bacterial flora include unsupervised learning algorithms and supervised learning algorithms.Due to the similarity between the characteristics of the microbial data and the characteristics of the text data,the Latent Dirichlet Allocation(LDA)probabilistic topic model is used in this paper.However,in the experimental processing,it is found that the LDA model has some defects,so on this basis,a model based on the minimized weighted edge value backbone tree is introduced to improve the Celltree-LDA probability topic model.Compared with the traditional method of clustering of intestinal flora,K-Means clustering,and LDA model,the validity of Celltree-LDA model was proved.Then,the results of Celltree-LDA model were analyzed,combined with clinical experiments,to further excavate the corresponding clinical significance and biological significance.(1)By using the folding Gibbs sampling algorithm in the Celltree-LDA model,the time heterogeneity Operational Taxonomic Unit(OTUs)data set of the first sets of data sources of the MVB and the minor hepatic encephalopathy(MHE1)bacteria were analyzed.The results show that Celltree-LDA model can distinguish the heterogeneity between samples,and it is more effective than LDA model,system clustering and K-Means clustering method in mining intestinal microflora structure.More importantly,the Celltree-LDA model also identifies the OTUs with the greatest impact on the sample.(2)According to the Celltree-LDA model analysis,the time-course heterogeneity of intestinal microbiota using rifaximin combined probiotics in patients with mild hepatic encephalopathy(MHE2)in the second set of data sources using the Celltree-LDA model The structure(OTUs population dataset)and function(KO Metabolic Dataset)were analyzed.The bacterial structure study showed that the Celltree-LDA model identified three OTUs that had the greatest heterogeneity in the structure of the gut microbiota.According to the changes in themicrobial flora within the group before and after treatment,the three species were found.Changes in the structure of the intestinal flora after treatment(P < 0.05).The study of bacterial flora showed that rifaximin did not alter the function of intestinal flora in patients with MHE,while rifaximin combined with probiotics altered the function of intestinal flora in MHE patients(P<0.05).In addition,according to the clinical efficacy index: comparing the serum inflammatory factors and blood ammonia levels after treatment between the two groups,the observation group was significantly better than the control group,with statistical significance(P<0.05).In summary,the Celltree-LDA model not only efficiently quantifies the heterogeneity of the bacterial structure and function,but also identifies the OTUs that have the greatest heterogeneity.Rifaximin combined with probiotic therapy can significantly improve blood ammonia levels and serum inflammatory factors in patients with MHE,and change the structure and function of intestinal microbiota in patients with MHE.It has good clinical reference value.