我的毕业设计是针对美国杜克大学医学院的一组实验研究数据进行的统计学分析。杜克大学医学院的学者通过研究发现，患有Ⅱ型糖尿病的患者再患心血管疾病的可能性很高。而阿司匹林是著名的预防抵抗心血管疾病的药物。但是，经验表明，患有Ⅱ型糖尿病的患者会对阿司匹林表现出一种“抵抗”作用。而这种作用的生理机制仍然是未知的。所以，为了探究这种生理机制，找出与之相关的基因，杜克大学医学院的研究人员设计了一系列生物学实验，得到了一组数据。而我的工作就是运用统计学工具R语言中的基本数据处理操作和相关建模及分析数据的函数包，来尝试寻找是否存在某种基因或者生物反应渠道，改变了人们对阿司匹林的反应效应，找出这些与之相关的基因或者机制，并且再在Ⅱ型糖尿病患者和健康人群之间比较其差别。这项分析分成两个部分。在第一部分中，我运用R语言中的t检验，首先检测了我从杜克大学医学院研究人员那里得到的原始待分析数据是否存在“批次效应”，即在生物实验设计时是否有效控制了变量，从而能使对数据的统计学分析有意义，结果可靠。然后应用R语言中用来处理生物统计数据的limma和GSEA分析方法找出显著表达的基因。在第二部分中，为了寻找目标因变量——pfsi和全部13029个被测基因之间的数学关系，找出最大相关度基因，我运用了贝叶斯变量选择的方法，以及lasso回归的方法，最后确定在全部13029个基因中，不存在对pfsi有显著影响的基因。由此说明，生物学和医学研究人员还需要进一步寻找新的复合变量，来取代pfsi，用以解释血小板的相关参数，再对其与基因的相关性进行进一步检测。

Patients with Type 2 Diabetes (T2D) are at an increased risk for cardiovascular disease (CVD). Aspirin (ASA) is well-known inhibitor of platelet function and prevents CVD, however, T2D patients appear to be “resistant” to the effects of aspirin. The molecular mechanisms that underlie the differential response in T2D patients are unknown. Using statistical analysis, we try to find if there are genes / networks / pathways that change in response to aspirin exposure, specify them and compare them between healthy (HV) and T2D cohorts to see if there is any difference. In the first part of our analysis, we examine the batch effect, using t test in R, and then apply limma method and GSEA to find out the statistically significantly expressed genes / pathways in both cohorts. In the second part, we try to fit a model to explain the relationship between the total 13029 genes and the dependent variable, pfsi and find out the most responsible genes.

[1] Voora D, Cyr D, Lucas J, Chi JT, Dungan J, McCaffrey TA, Katz R, Newby LK, Kraus WE, Becker RC, Ortel TL, Ginsburg GS. (2013) Aspirin exposure reveals novel genes associated with platelet function and cardiovascular events. J Am Coll Cardiol. 2013 Oct 1;62(14):1267-76. doi: 10.1016/j.jacc.2013.05.073. Epub 2013 Jul 3.

[2] Voora D, Ortel TL, Lucas JE, Chi JT, Becker RC, Ginsburg GS. (2012) Time-dependent changes in non-COX-1-dependent platelet function with daily aspirin therapy. J Thromb Thrombolysis. 2012 Apr;33(3):246-57.doi: 10.1007/s11239-012-0683-0

[3] Yee DL, Sun CW, Bergeron AL, Dong JF, Bray PF (2005) Aggregometry detects platelet hyperreactivity in healthy indi- viduals. Blood 106(8):2723–2729

[4] FitzGerald GA, Oates JA, Hawiger J, Maas RL, Roberts LJ II, Lawson JA, Brash AR (1983) Endogenous biosynthesis of pros- tacyclin and thromboxane and platelet function during chronic administration of aspirin in man. J Clin Invest 71(3):676–688

[5] Pulcinelli FM, Pignatelli P, Celestini A, Riondino S, Gazzaniga PP, Violi F (2004) Inhibition of platelet aggregation by aspirin progressively decreases in long-term treated patients. J Am Coll Cardiol 43(6):979–984

[6] Perneby C, Wallen NH, Rooney C, Fitzgerald D, Hjemdahl P (2006) Dose- and time-dependent antiplatelet effects of aspirin. Thromb Haemost 95(4):652–658

[7] Subramanian, Tamayo, et al. (2005, PNAS 102, 15545-15550) and Mootha, Lindgren, et al. (2003, Nat Genet 34, 267-273).

[8] Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W and Smyth GK (2015). “limma powers differential expression analyses for RNA-sequencing and microarray studies.” Nucleic Acids Research, 43(7), pp. e47.

[9] 人过留名. lasso思想及算法 http://liuzg202.blog.163.com/blog/static/29427196201010259233251/. 2016年4月18日访问.

[10] Jun Liu. Large-Scale Sparse Logistic Regression[J]．KDD’09．

[11] Wang Zhanfeng. A LASSO-Type Approach to Variable Selection and Estimation for Censored Regression Model[J]．2010,02．

[12] 写长城的诗.用glmnet包实施套索算法（lasso）http://www.r-bloggers.com/lang/chinese/1017. 2016年4月22日访问.