Table_1_Identification of Microbiota Biomarkers With Orthologous Gene Annotation for Type 2 Diabetes.XLSX

Type 2 diabetes (T2D) is a systematic chronic metabolic condition with abnormal sugar metabolism dysfunction, and its complications are the most harmful to human beings and may be life-threatening after long-term durations. Considering the high incidence and severity at late stage, researchers have been focusing on the identification of specific biomarkers and potential drug targets for T2D at the genomic, epigenomic, and transcriptomic levels. Microbes participate in the pathogenesis of multiple metabolic diseases including diabetes. However, the related studies are still non-systematic and lack the functional exploration on identified microbes. To fill this gap between gut microbiome and diabetes study, we first introduced eggNOG database and KEGG ORTHOLOGY (KO) database for orthologous (protein/gene) annotation of microbiota. Two datasets with these annotations were employed, which were analyzed by multiple machine-learning models for identifying significant microbiota biomarkers of T2D. The powerful feature selection method, Max-Relevance and Min-Redundancy (mRMR), was first applied to the datasets, resulting in a feature list for each dataset. Then, the list was fed into the incremental feature selection (IFS), incorporating support vector machine (SVM) as the classification algorithm, to extract essential annotations and build efficient classifiers. This study not only revealed potential pathological factors for diabetes at the microbiome level but also provided us new candidates for drug development against diabetes.

2型糖尿病（Type 2 diabetes, T2D）是一类以糖代谢紊乱为核心的系统性慢性代谢性疾病，长期病程后其并发症对人体危害最甚，甚至可危及生命。鉴于其高发病率与晚期病情的严重程度，研究者一直致力于在基因组、表观基因组与转录组层面，识别2型糖尿病的特异性生物标志物及潜在药物靶点。微生物参与包括糖尿病在内的多种代谢性疾病的发病进程，但目前相关研究仍缺乏系统性，且未对已识别的微生物开展功能层面的探索。为填补肠道微生物组与糖尿病研究间的这一研究空白，本研究首先引入eggNOG数据库与KEGG同源基因（KEGG ORTHOLOGY, KO）数据库，对微生物群进行同源（蛋白质/基因）注释。本研究采用了两份带有此类注释的数据集，并通过多种机器学习模型开展分析，以识别2型糖尿病的关键微生物群生物标志物。本研究首先采用极具实用性的特征选择方法——最大相关最小冗余（Max-Relevance and Min-Redundancy, mRMR）对两份数据集进行处理，为每份数据集生成特征列表。随后将该特征列表输入至增量特征选择（Incremental Feature Selection, IFS）流程中，并以支持向量机（Support Vector Machine, SVM）作为分类算法，以提取核心注释信息并构建高效分类器。本研究不仅在微生物组层面揭示了糖尿病的潜在致病因素，同时也为抗糖尿病药物研发提供了全新的候选靶点。