| ObjectiveA large number of studies have shown that the mechanism related to intestinal flora is important in the occurrence and development of coronary heart disease.However,due to the large variety,large number and complex role of intestinal flora,it is still not clear what kind of intestinal flora is related to coronary heart disease CHD.Therefore,this study aims to mine the characteristic gut microbiota of patients with CHD based on machine learning methods,and on this basis,to establish a mouse model of CHD and use Tan-Yu Tongzhi Prescription(TYTZ)to clarify the effect of prescription on the regulation of gut microbiota.Methods1.Mining of Coronary Heart Disease Characteristic Microbiota Based on Multi-model Data FusionThe sequencing data of the gut microbiota in patients with CHD were retrieved from the National Center for Biotechnology Information,Gut Microbiota Data Repository,National Genome Science Data Center of China,European Nucleotide Archive database and other databases,which will be available.FASTQ file for the project data download.QIIME 2 software was used for data analysis and species annotation,and then random forest classifier,gradient boosting decision tree classifier and AdaBoost classifier were constructed to predict the characteristic flora.Two categories were used for grouping,80%were selected as the training set and 20%as the test set,5-fold cross validation was used to output the scores of important characteristic variables.At the same time,the SHAP addition and explanation feature contribution evaluation algorithm is used to evaluate the results of the machine learning method model.Finally,the species importance of machine learning prediction results and the evaluation results of SHAP feature contribution were combined to propose a method to evaluate the importance index V of characteristic species,and the coronary heart disease characteristic flora was determined by importance ranking.2.Construction of Coronary Heart Disease Model Mice and Study on Intestinal Flora CharacteristicsThere were 15 SPF male C57BL/6J mice and 15 ApoE-/-mice in sham operation group and coronary heart disease model group,each group was fed continuously for 12 weeks.The model group was fed with high fat diet throughout the whole process,and the animal model of coronary heart disease was constructed by ligation of left anterior descending branch coronary artery at the 8th week.At the end of the feeding period,Left ventricular ejection Fraction(EF),left ventricular Fractional Shortening(FS),Stroke Volume(SV),left ventricular mass(LVM),cardiac output(cardiac output,CO),Left ventricular anterior wall systolic thickness(Left ventricular anterior wall systolic thickness,LVAWs),Left ventricular anterior wall diastolic thickness(Left ventricular anterior wall diastolic thickness,LVAWd),Left ventricular posterior wall systolic thickness(Left ventricular posterior wall systolic thickness,LVPWs),Lft ventricular posterior wall diastolic thickness(LVPWd)and left ventricular wall thicknesswere measured by echocardiography.Serum samples were collected for Total Cholesterol(TC),High density lipoprotein cholesterol(HDL-C),High density lipoprotein cholesterol(HDL-C),Low density lipoprotein cholesterol(LDL-C)and Triglyceride(TG)were detected.Aorta was collected for oil red O staining,and feces were collected for metagenome sequencing.3.TYTZ on Intestinal Flora in Mice with Coronary Heart DiseaseFifteen SPF male C57BL/6J mice were used as the sham operation group,and 30 ApoE-/-mice were randomly divided into model group and TYTZ group.The mice were fed with high-fat diet or normal diet for 12 weeks.At the fourth week,the coronary heart disease animal model was established by ligation of the left anterior descending coronary artery.At the end of the feeding period,the mice in each group were examined by echocardiography,and the serum was collected for TG,TC,LDL-C and HDL-C and non-targeted metabolomics based on ultra-high performance liquid chromatography-mass spectrometry.The aorta was collected for oil red O staining,and the feces were collected for 16S rRNA sequencing of intestinal flora to observe the regulatory effect of TYTZ on the intestinal flora of CHD model mice.Results1.A total of 493 patients with coronary heart disease were included in the study,which were divided into single-end data and double-end data according to different sequencing platforms.After the integration of heterogeneous data,different methods of machine learning prediction analysis were performed.The AUC of the single-end data and the double-end data of the random forest screening model were 0.84 and 0.95,the ACC were 0.85 and 0.96.A total of 72(42 and 30)differential species were screen with scores of important characteristic variables of output greater than 0.01.The AUC of the gradient boosting classifier algorithm was 0.94 and 0.92,ACC were 0.92 and 0.86;A total of 27(14 and 13)differential species were screen with scores of important characteristic variables of output greater than 0.01.The AUC of the AdaBoost classification method were 0.81 and 0.74,ACC were 0.86 and 0.81.A total of 27 and 29 differential species with scores of important characteristic variables of output greater than 0.01 were obtained.After SHAP evaluation on the results of three machine learning algorithms,combining the results of machine learning algorithm and SHAP evaluation results,the average value of the importance index V of the evaluated characteristic species was 17.92,and the median value was 12.18786.The species larger than the median value were selected as a total of 47 characteristic flora with high probability of coronary heart disease.2.The results of echocardiography showed that the EF,FS,SV,and CO values of the model group were significantly lower than those of the sham operation group,and the LVAWd and LVPWs values were significantly increased(P<0.05),and the left ventricular anterior wall was significantly thinner,suggesting that myocardial function was damaged and myocardial contractility was weakened.The results of blood lipid showed that the contents of TC,TG and LDL-C in the model group were significantly increased,and the content of HDL-C was significantly decreased(P<0.05).Oil red O staining showed that there was more aortic plaque formation in the model group,but almost no plaque formation in the sham operation group.Metagenomics results showed that the sample quality of the two groups was qualified,and there were significant differences in Alpha diversity and Beta diversity(P<0.05).Combined with the screening results in the first part,finally,7 characteristic bacterial flora that both functioned in patients with coronary heart disease and animal models were identified.They were Enterobacteriaceae,Enterococcaceae,Barnesiellaceae,Megamonas,Akkermansiamuciniphila,Prevotellacopri,Holdemania.3.The results of echocardiography showed that EF,FS,CO and SV in the model group were significantly lower than those in the sham operation group,while LV Mass,LVAWd and LVPWd were increased significantly.After the treatment of TYTZ,EF,FS,CO,SV increased significantly,LV Mass,LVAWd,LVPWd decreased significantly.Compared with the sham-operation group,TC,TG and LDL-C in the model group increased significantly,while HDL-C decreased significantly.Compared with the model group,TC,TG,LDL-C in the TYTZF group decreased significantly(P<0.01 or P<0.05).Ultrasonography showed that the left ventricular anterior wall of the mice in the model group was significantly thinner,and the degree of left ventricular anterior wall thinning was reduced after the treatment of TYTZ.The oil red O staining results showed that almost no lipid tissue was stained in the aortic vessels of the sham operation group,and there were scattered colored areas everywhere in the inner wall of the aortic vessels in the model group,especially in the aortic arch,with obvious lipid deposition.In the TYTZ group,a small amount of scattered staining areas were observed in the inner wall of the aorta,and the staining in the aortic arch was lighter,and the lipid deposition in the blood vessel was less than that in the model group.Based on the results of intestinal flora of all parts,TYTZ can regulate the three core flora of FirmicutesbacteriumM102,Faecalibaculum,Megamonas,in the mice model of CHD and the six characteristic species of Coriobacteriaceae、Collinsella、Parabacteroides、Megamonas、Streptococcus、Anaerofustis、Parabacteroides distasonis in patients with CHD.Megamonas has species in both patients with coronary heart disease and the mice model,which may be the core characteristic flora of TYTZ.Metabonomics results showed that 528 metabolites were changed after treatment with TYTZ,which were related to galactose metabolism,terpenoid skeleton biosynthesis,fructose and mannose metabolism,cholesterol metabolism pathway,monocbacteriocin biosynthesis and metabolism pathway.Ileibacterium,unclassifiedRhodospirillales and Faecalibaculum had the most related metabolites.Megamonas,Tyrosyl Alanine,Isovaleryl glucuronide and 2-Hydroxy-desipramine glucuronide),Saccharopine,Lyso-PC and other metabolites were positively correlated.The core characteristic flora of TYTZ.Conclusion1.Based on the sequencing data of clinical intestinal flora of patients with CHD in the public database,three machine learning models and SHAP model interpretation algorithm were integrated,and a feature importance evaluation index V value was proposed.Finally,93 possible characteristic intestinal flora of CHD were screened out,and the 47 most probable characteristic flora were those greater than the median score.2.High-fat feeding combined with left anterior descending coronary artery ligation can establish a mouse model of CHD.After comparing the results of fecal metagenomics with the characteristic flora of CHD predicted in the first part,7 microflora were obtained in both parts:Enterobacteriaceae,Enterococcaceae,Barnesiellaceae,Megamonas,Akkermansiamuciniphila,Prevotellacopri and Holdemania,which act together in animals and patients with CHD,and can be used as a characteristic flora of CHD for research.3.The TYTZ can improve cardiac EF,FS,CO,SV,SV,LV Mass,LVAWd,LVPWd,reduce TC,LDL-C,TG,and TYTZ regulated the CHD mouse model and the CHD patient Faecalibaculum,Megamonas,FirmicutesbacteriumM102,Coriobacteriaceae,Collinsella,Parabacteroides,Megamonas,Streptococcus,AnaerofustisParabacteroides distasonis and Megamonas,Among them,Megamonas may be the core species of TYTZ.The regulatory effects of TYTZ are related to galactose metabolism,terpenoid skeleton biosynthesis,fructose and mannose metabolism,cholesterol metabolism pathway,and monocbacteriocin biosynthesis and metabolism pathway.In addition,it can also regulates Tyrosyl Alanine,Isovaleryl glucuronide,Saccharopine,2-Hydroxy-desipramine glucuronide and Lyso-PC metabolic pathways by affecting Megamonas. |