A Study on Metagenomic Analysis and Feature Selection of the Subgingival Microbiota Associated with Periodontal Disease Using Machine Learning Algorithms

• Main Authors: CHEN, WEN-PEI, 陳文培
• Other Authors: TASI, SUH-JEN
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/jhfnq7

博士 === 靜宜大學 === 應用化學系 === 107 === Periodontitis is an inflammatory disease involving complex interactions between oral microorganisms and the host immune response. It was reported as a disease associated with alteration of the subgingival microbiota. Although it can be treated by mechanical nonsurgical therapy, the disease may recur and progress intermittently. Understanding the structure of the microbiota community associated with periodontitis is essential for improving classifications and diagnoses of various types of periodontal diseases and will facilitate clinical decision-making. In this study, we integrated the open-source softwares and proposed a 16S rRNA metagenomic analysis platform. Using this platform, we investigated and compared the compositions of the microbiota communities of subgingival plagues samples from healthy individuals and patients with periodontitis. We identified the core microbiota and observed differences between health- and periodontitis-associated bacterial communities at all phylogenetic levels. We discovered that the genera Porphyromonas, Treponema, Tannerella, Filifactor, and Aggregatibacter were more abundant in patients with periodontal disease, whereas Streptococcus, Haemophilus, Capnocytophaga, Gemella, Campylobacter, and Granulicatella were found at higher levels in healthy controls. We also performed a 6-month follow-up study to characterize the dynamic changes in the subgingival microbiome of periodontitis patients before and after nonsurgical periodontal treatment of the same teeth. We aimed to characterize periodontitis-associated microorganisms ahead of the appearance of the clinical symptoms. The therapy resulted in a greater influence on the microbiome 3 months after therapy, but it was gradually recovered at 6 months even though the clinical parameters for a full-mouth clinical examination were stable up to 6 months post-treatment. Our finding suggested that there was a significant relationship between clinical parameters and common periodontal pathogens, including Porphyromonas gingivalis, Filifactor alocis, Treponema denticola and Tannerella forsythia. There were 15 bacteria, including Acinetobacter baumannii, Streptococcus sanguinis, Campylobacter gracilis, Corynebacterium matruchotii, Capnocytophaga granulosa, and Capnocytophaga sputigena, which increased abundance after treatment. On the other hand, there were 9 bacteria, including famous periodontitis-associated species Porphyromonas gingivalis, Tannerella forsythia, Treponema denticola, Filifactor alocis, and Treponema socranskii, which reduced abundance after treatment. Notably, most of these bacteria have a tendency to revert to the pre-treatment state. Furthermore, we proposed a novel feature selection algorithm for selecting features with more information from many variables. The combination of these features and machine learning methods was used to construct prediction models for predicting the health status of patients with periodontal disease. Using the feature selection algorithm, prediction models could accurately predict the health status of samples by examining fewer features.