減少檢定個數以快速檢測基因之間交互作用效應的分析方法

Abstract

許多研究顯示，複雜疾病不只是受到單一基因的影響，而是受到多個基因同時進行調控。隨著微陣列晶片技術的進步，目前最多可同時偵測數百萬個單一核苷酸多型性(single nucleotide polymorphism, SNPs)；若欲考慮這些SNPs之間的交互作用與疾病的關係時，需要龐大的計算量。為了解決這樣的問題，本論文提出高效率偵測法(Efficient Detecting Method)，利用卡方檢定的特性，有效減少需要檢定的SNPs交互作用數，進而降低運算所需的時間。同時，分別根據卡方檢定P值的排序與模式選擇，比較高效率偵測法與傳統上檢定所有交互作用的結果。此論文分析了三筆包含不同數目SNPs的實際資料，由結果看來，高效率偵測法都能有效的降低運算量，同時檢測出顯著與疾病相關的SNPs交互作用。

Previous studies have shown that complex diseases are affected by multiple genes. With the development of microarray technique, million single nucleotide polymorphisms (SNPs) could be genotyped for association analyses simultaneously. There will be many possible combinations of SNP-SNP interactions, therefore a great amount of calculation is necessary. We proposed the efficient detecting method (EDM) to reduce the number of association tests and decrease the calculation time by chi-square test. The testing p-values and model selection procedures were utilized for comparing the testing results of EDM and all combinations. Three datasets with different numbers of SNPs were analyzed to demonstrate the performance of EDM. From the results, we concluded that EDM could decrease the calculation time and detect the significant SNP-SNP interactions associated with diseases.