·糖尿病分子生物学与遗传学·

　　杜文聪 陆莹 叶新华 李倩 俞晓芳 成金罗 马建华 高燕勤 杜娟 石慧 曹园园 周玲

　　基金项目：国家自然科学基金资助项目(30771858);江苏省自然科学基金资助项目(BK2007229)

　　作者单位：210029南京，南京医科大学公共卫生学院流行病与卫生统计学系(杜文聪、陆莹、杜娟、石慧、曹园园、周玲);南京医科大学附属常州第二人民医院(叶新华、成金罗);南京市第一医院内分泌科(李倩、马建华);南京医科大学第三附属医院内分泌科(高燕勤、俞晓芳)

　　通讯作者：周玲，E-mail：lzhou@njmu.edu.cn

　　【摘要】目的探讨BP人工神经网络(BPANN)在脂联素(APN)基因单核苷酸多态性(SNP)与中国南方地区汉族人群T2DM易感性关系中的应用特点。方法采用BPANN分析方法， 对影响因素按照平均影响值(MIV)的绝对值大小排序，并与Logistic回归模型及单倍型分析的结果相比较。结果BPANN模型显示，在T2DM相关危险/保护因子中，相关危险因子为腰围、rs12495941、高血压史、T2DM家族史、rs266729、性别、吸烟、年龄、rs16861194、rs16861205、rs1063539、BMI、高脂血症史、rs2241767、rs7649121、rs3821799、rs822394;相关保护因子为血清APN浓度、饮酒、rs182052。MIV位于前3位的为血清APN浓度、腰围、rs12495941。多因素Logistic回归分析中因子顺位为T2DM家族史、饮酒、高血压史等。将BPANN多因素分析中MIV位于前10位的3个SNPs位点进行单倍型分析，结果表明，与最常见的单倍型TGC相比，rs266729G的单倍型TGG可增加T2DM的患病风险，其OR(95%CI)为2.42(1.17~4.99)。结论APN基因多态性位点间的交互作用与T2DM易感性存在关联。BPANN用于筛选T2DM等复杂多病因疾病的影响因素，可能具有一定的优势。

　　【关键词】BP人工神经网络;糖尿病,2型;脂联素

　　doi:10.3969/j.issn.1006-6187.2012.01.008

　　Association between adiponectin (APN) gene polymorphism locus interacts and type 2 diabetes risk in a Chinese Han population studied by BPANNDU Wen-cong, LU Ying, YE Xin-hua,et al.Department of Epidemiology and Biostatistics, School of Public Health, Nanjing Medical University, Nanjing 210029， China

　　Corresponding author： ZHOU Ling, E-mail：lzhou@njmu.edu.cn

　　【Abstract】ObjectiveTo explore the association between APN gene polymorphism locus interacts and type 2 diabetes risk in a Chinese Han population studied by back-propagation artificial neural network (BPANN). MethodsBy BPANN, the mean impact value (MIV) of each input variables was sequenced and was compared with logistic regression model and haplotype analysis. ResultsBy BPANN analysis, the risk factors were waist, rs12495941, hypertension, diabetes mellitus family history, rs266729, gender, smoking, age, rs16861194, rs16861205, rs1063539, BMI, hyperlipidemia, rs2241767, rs7649121, rs3821799, rs822394,and the protective factors were serum adiponectin level, drinking, rs182052. The importance order for the factors of diabetes mellitus was serum adiponectin level, waist, rs12495941, et al. In multiple logistic regression analysis, the importance order appeared to be diabetes mellitus family history, not alcohol drinking, hypertension, et al. Haplotype TGG, with variant G of rs266729, could increase the risk for T2DM as compared with the most common haplotype TGC[OR (95%CI)=2.42(1.17-4.99)]. ConclusionThe APN gene polymorphisms locus interact is associated with the susceptibility of T2DM. Screening by BPANN for factors of T2DM and other complex factors causing diseases, may have certain advantages.

　　【Key words 】Back-propaqation artificial neural network; Diabetes mellitus, type 2; Adiponectin