| With the continuous development of dental implant technology,implant therapy had become one of the preferred methods for partially edentulous patients.However,post-implant complications such as peri-implant disease had a great negative impact on the efficacy of implant therapy.Plaque was the primary cause of peri-implant disease[1],according to the available research,but the pathogenesis of peri-implant disease was still unclear,so it was difficult to establish targeted regulatory strategies.In this study,information and biological samples were collected from the prospective cohort of implant patients,and microbiome and metabolomics techniques were used to identify risk markers of peri-implant disease and explore possible metabolic pathways.Moreover,by combining clinical factors,statistical analysis and optimization were carried out to generate a clinical assessment model of the risk of peri-implant disease.This study aimed to develop a theoretical framework and basis for the early prevention and treatment of peri-implant diseases,and to provide a potent tool for clinical education and behavioral intervention.[Aim]In this study,metabolomics and microbiome analysis methods were used to pinpoint the distinct metabolites and metabolic pathways of peri-implant diseases by analyzing the variations of peri-implant crevicular fluid and saliva metabolites in different health conditions around implants.It then studied the relationship between peri-implant subgingival plaque and peri-implant crevicular fluid metabolites order to investigate the pathogenesis of peri-implant disease.Lastly,different machine learning methods were used to construct a risk assessment model of peri-implant disease combined with baseline clinical factors and metabolite characteristics.This study would provide scientific support and theoretical basis for the development of risk management measures for peri-implant disease,in order to further reduce the incidence of peri-implant disease and optimize the therapeutic effect of implantation.[Methods]1.Study on the association between peri-implant gingival crevicular fluid and saliva metabolites and peri-implant disease.The prospective group of patients underwent implant restoration at the Stomatology Hospital of the Fourth Military Medical University’s prosthodontics department from June2014 to September 2015 was followed up.Oral examination and questionnaire survey were performed on the patients,study samples were collected.Samples of peri-implant gingival crevicular fluid and saliva were analyzed by ultra performance liquid chromatography tandem mass spectrometry（UPLC-MS）.To explore the distribution of peri-implant gingival crevicular fluid,salivary metabolites and metabolites of peri-implant different health status,principal component analysis（PCA）,orthogonal partial least squares discriminant analysis（OPLS-DA）,partial least squares discriminant analysis（PLS-DA）were used.Statistical methods,such as OPLS-DA and Kyoto Encyclopedia of Genes and Genomes（KEGG）pathway enrichment,were used to study the differences in metabolites and metabolic pathways in different health conditions of implants.The metabolites with variable important in projection（VIP）of OPLS-DA greater than 1 and Student’s t test P less than 0.05 were considered as differential metabolites.When P value was less than 0.05in KEGG enrichment analysis,the pathway was considered to be significantly enriched.2.Study on the association between peri-implant microorganisms and gingival crevicular fluid metabolites and metabolic pathwaysSubgingival plaque of 35 enrolled patients was analyzed by 16S r RNA microbiome,followed by a combined metabonomics and microbiome analysis.Hierarchical clustering method was used to reduce the dimension of metabolites distribution,and Pearson correlation analysis was used to explore the association between metabolites and microorganisms.Pearson correlation analysis was conducted to study the correlation between different metabolites and different metabolites in different health states around the implants.The co-occurrence probability analysis of metabolites and microflora was conducted to predict the occurrence probability of corresponding metabolites in the presence of certain microorganisms,and further determine the interaction relationship between microorganisms and microflora.KEGG pathway enrichment analysis was performed on microorganisms and differential metabolites to study the differential metabolic pathways in different health states of implants.3.Establish a risk model of peri-implant disease based on metabolites and common risk factorsThe data matrix of 33 baseline clinical factors of patients and 166 differential metabolites of gingival crevicular fluid was constructed,the data were preprocessed,the metabolic data was changed into binary variables,and then variable screening was conducted.The metabolites were screened from VIP of OPLS-DA,P value of Student’s t test,Fold Change（FC）and area under curve（AUC）of receiver operating characteristic curve（ROC）.Clinical factors were screened from Spearman correlation analysis and previous research experience,and finally a modeling data matrix of 30 metabolic factors and 20 clinical factors in 35 samples was constructed.Based on R software,the decision tree,random forest and support vector machine models of clinical factors and metabolic factors in different health states around the implant were constructed.At the same time,three models containing only 20 clinical factors were constructed for comparison.Four indexes AUC,precision ratio,recall ratio and F1 value were used to evaluate the model performance.[Results]1.There were differences in gingival crevicular fluid metabolites and metabolic pathways in different health conditions around implantsUPLC-MS analysis was performed on gingival crevicular fluid and saliva samples from 36 patients in the prospective cohort who met the inclusion and exclusion criteria（12patients in the peri-implantitis group,12 patients in the peri-implant mucositis group,and12 patients in the peri-implant health group）.PCA,PLS-DA and OPLS-DA analysis suggested that there was a certain difference between gingival crevicular fluid and saliva metabolites（PLS-DA Q2cum=0.986）.There was a certain difference in the distribution of gingival crevicular fluid metabolism among the three groups,but no significant difference in the distribution of metabolites in saliva samples.The differential metabolites of gingival crevicular fluid were analyzed.There were 138 kinds of differential metabolites between peri-implantitis group and peri-i mplant health group.The top 3 metabolites of VIP value were threonine-phenylalan ine-leucine-glutamine（VIP=3.957）,5-methoxytryptophan（VIP=3.829）,isoleucine-glut amine-alanine-valine（VIP=3.728）.There were 171 different metabolites between the peri-implant mucositis group and the peri-implant health group.The top 3 metabol ites with VIP value were glutamylglutamine（VIP=4.474）,γ-glutamylornithine（VIP=4.288）and leukotriene E4（VIP=4.141）.There were 75 different metabolites betwee n the peri-implantitis and the peri-implant-mucitis group.The top 3 metabolites wit h VIP value were phenylalanine proline（VIP=3.426）,prostaglandin E1（VIP=3.406）and inosine 5’-monophosphate（VIP=3.357）.The differential metabolic pathway of gingival crevicular fluid was analyzed,enrichment of arginine biosynthesis（P=0.002）and histidine metabolism pathway（P=0.004）were the highest in peri-implantitis and peri-implant health groups.In the peri-implant mucositis and peri-implant health groups,lysine degradation（P<0.001）,purine metabolism（P<0.001）and rginine biosynthesis（P=0.001）were the most significant ways of enrichment.In the peri-implantitis and peri-implant mucositis groups,the most significant enrichment pathways were purine metabolism（P<0.001）,biosynthesis of cofactors（P=0.001）,arginine and proline metabolism（P=0.004）.2.There was a certain correlation between peri-implant microbiota and gingival crevicular fluid metabolites and metabolic pathwaysA total of 35 samples were included in this study,and 1 case was obviously eliminated due to outliers in PCA and PLS-DA analysis.The co-occurrence probability analysis of microbial and gingival crevicular fluid metabolites in 35 samples showed that Eikenella,Selenomonadaceae,Lautropia had strong co-occurrence relationship with a variety of different metabolites.Correlation analysis showed that 47 groups of metabolites were correlated with bacterial community in the peri-implantitis group,and the most significant correlation was 27 groups of metabolites and Armatimonadota（R=0.979,P<0.001）;In the peri-implant mucositis group,a total of 66 groups of metabolites were correlated with the bacterial community,and the most significant correlation was 23 metabolites and Caldisericota（R=0.944,P<0.001）;A total of 87 groups of metabolites were correlated with the flora in the peri-implant health group,and the most significant correlations were metabolite clusters 7 and GAL15（R=0.913,P<0.001）.In the peri-implantitis group and peri-implant health group,there were a total of 126different metabolites and different species combinations were significant（P<0.05,|r|>0.3）,and there were 122 highly correlated combinations,the correlation analysis between Prolylhydroxyproline and solirubrobacter（r=-0.772,P<0.001）had a significant highest correlation.In the peri-implant mucositis group and the peri-implant health group,there were 318 kinds of correlation between metabolites and bacterial in pin-two analysis,and257 kinds of highly correlated combinations,PC（16:0/18:0）had the strongest correlation with Komagataeibacter（r=-0.770,P<0.001）.In the peri-implantitis group and peri-implant mucositis group,35 combinations were correlated,and 18 combinations were highly correlated.The correlation between dethiobiotin and Roseiflexaceae was the strongest（r=0.798,P<0.001）.There were significant differences in metabolic pathways such as Th17 cell differentiation（Abundance=2.065）,estrogen signaling pathway（Abundance=2.064）between the peri-implantitis group and the peri-implant health group.There were significant differences in thiamine metabolism（Abundance=1.917）,beta-Alanine metabolism（Abundance=1.821）between the peri-implant mucositis and peri-implant health groups.In the enrichment analysis of pathways between the peri-implant mucositis group and the peri-implantitis group,the most significant differences were in benzoate degradation（Abundance=2.275）and Th17 cell differentiation（Abundance=2.168）.3.The peri-implant disease risk model based on metabolites and common risk factors had good performanceThe 166 kinds of differential metabolites of periimplant-related diseases were transformed into dichotomous variables,20 kinds of differential metabolites with OPLS-DA VIP greater than 3 and T-test P less than 0.01 were selected,3 kinds of differential metabolites with FC>3 or FC<1/3 and 10 kinds of metabolites with the highest AUC value ROC were added.A total of 30 metabolite factors were formed.Among the 32 clinical factors,3 were significant in correlation analysis,and 17 were added in multivariate and univariate analysis to form 20 clinical factorsThe data matrix was divided into a training set and a test set according to 6:4.The machine learning methods of decision tree,random forest and support vector machine were used to build a risk model of periimplant-implant disease based on metabolites and common high-risk factors.In the test set,the AUC of support vector machine model and random forest model in the test set reached 0.933 and 1.000,showing good predictive performance.The AUCs of clinical factor decision tree model,metabolite and clinical factor decision tree model,clinical factor random forest model,metabolite and clinical factor random forest model,clinical factor support vector machine model,metabolite and clinical factor support vector machine model are:0.625,0.689,0.771,1.000,0.667,0.956;Compared with the prediction model based solely on clinical factors,the model combined with metabolic factors has better predictive performance.Further optimization of the model showed that the AUC of the random forest model composed of threonine-asparagine-valine-leucine and isoleucine was 0.988,showing good predictive performance.[Conclusion]1.Peri-implant gingival crevicular fluid metabolites could distinguish peri-implantitis,peri-implant mucositis and peri-implant health population,and the screened differential metabolites may be metabolic markers of peri-implant disease.2.There were significant differences in metabolic pathways such as arginine biosynthesis,histidine metabolism,lysine degradation,and purine metabolism among different health states around the implant,which may be involved in the pathogenesis and progression of the peri-implant disease.3.Armatimonadota,Caldisericota and other bacteria may exert a certain influence on the occurrence of peri-implant diseases by affecting the distribution of metabolites such as threonine-phenylalanine-leucine-glutamine.4.The risk assessment model of peri-implant disease constructed in this study had good predictive ability,especially the random forest model,the optimized model composed of two metabolites also has good predictive performance. |
PDF Full Text Request