论文部分内容阅读
目的:探讨BP人工神经网络(BPANN)在研究过氧化物酶体增殖物激活受体γ(PPAR-γ)和视黄醛α受体(RXR-α)基因单核苷酸多态性(SNP)位点与中国南方地区汉族人群2型糖尿病(T2DM)易感性关系中的应用特点。方法:采用BPANN分析方法,对591例2型糖尿病患者和724例正常对照者的基因多态性位点的分型结果、血清脂联素水平以及其他所有可能的影响因素按照平均影响值(MIV)的绝对值大小排序,并与Logistic回归模型的分析结果相比较,用多因子降维法(MDR)分析基因间的交互作用。结果:BPANN多因素分析中,2型糖尿病危险因子的顺位为血清脂联素浓度、高血压史、腰围、rs4240711、rs3132291、rs3856806、2型糖尿病家族史、饮酒、高脂血症史、吸烟、年龄、BMI指数、rs1045570、性别、rs2920502、rs6537944、rs4842194、rs17827276、rs1801282;而多因素Logistic回归分析中只有8个变量入选最终模型,因子顺位为高血压史、T2DM家族史、腰围、饮酒、吸烟、rs4240711、rs4842194、血清脂联素浓度;多因子降维法(MDR)分析结果显示模型X1X2X3(rs3856806,rs3132291,rs4240711)为最佳模型(交叉验证一致性10/10,P=0.0107)。结论:PPAR-γ和RXR-α基因多态性改变的交互作用对于中国南方汉族T2DM遗传易感性可能具有一定的作用。BPANN用于筛选T2DM等复杂多病因疾病的影响因素,可能提供更切合实际情况的模型。
OBJECTIVE: To investigate the effect of BP artificial neural network (BPANN) on single nucleotide polymorphism (SNP) of peroxisome proliferator activated receptor γ (PPAR-γ) and retinaldehyde α receptor (RXR-α) ) And susceptibility to type 2 diabetes mellitus (T2DM) in Han population of southern China. Methods: BPANN analysis was used to analyze the genotyping results, the level of serum adiponectin, and all other possible influencing factors in 591 type 2 diabetic patients and 724 normal controls according to the mean influence value (MIV ), And compared with the results of Logistic regression model, the interaction between genes was analyzed by multi-factor dimensionality reduction (MDR). Results: In BPANN multivariate analysis, the rank of risk factors for type 2 diabetes were serum adiponectin concentration, history of hypertension, waist circumference, family history of rs4240711, rs3132291, rs3856806, type 2 diabetes, history of drinking, hyperlipidemia, smoking , Age, BMI index, rs1045570, sex, rs2920502, rs6537944, rs4842194, rs17827276, rs1801282; only 8 variables were selected into the final model in multivariate Logistic regression analysis. The factors ranked as history of hypertension, family history of T2DM, waist circumference, alcohol consumption , Smoking, rs4240711, rs4842194 and serum adiponectin concentrations. The results of MDR analysis showed that the model X1X2X3 (rs3856806, rs3132291, rs4240711) was the best model (cross validation 10/10, P = 0.0107) . CONCLUSION: The interaction between polymorphisms of PPAR-γ and RXR-α may play a role in genetic susceptibility to T2DM in Han Chinese in southern China. The use of BPANN to screen for factors affecting complex polygenic disease, such as T2DM, may provide a more realistic model.