自從2003年首次發表後，以基因組為基礎的分析方法(GSA)在基因表達微陣列已經被廣泛地運用來探索基於網路知識之生物功能表現型的相關性。GSA專注在一組相關基因上，且相對於獨立基因分析(IGA)來說有更多優點。其中包括了更好的準確性，強韌性以及生物關聯。但以往的GSA研究並沒有考慮會抑制功能的基因組之間的關係。因此我們提出一個分析方案「以基因組為基礎的局部階層式叢集」(GSLHC)。此方法可以提供從功能，藥物反應等反應的生物見解。我們成功的應用GSLHC從各種基因的分子特徵資料庫(MsigDB)中製作出了C-Map。GSLHC可以消除細胞株本身對於基因表現的影響。從IGA分析的結果證明細胞株的影響明顯優於樣本類型以及藥物標靶。此外，GSLHC根據最顯著的基因集確定了18種功能藥物最相關的分子，其中8種包含公認的抗癌藥物。因此，GSLHC將有助於了解藥物對於生物體的影響，重新定位藥物，常見疾病的基因組診斷，以及功能為基礎的類異構癌症亞型分類模式診斷。Gene set-based analysis (GSA) has been widely utilized on gene expression microarray to explore the association of biological features with phenotypes based on a prior pathway knowledge since its first application in 2003. GSA focuses on sets of related genes and has exhibited major advantages over on individual gene analysis (IGA) with respect to greater accuracy, robustness, and biological relevance. However, previous GSA studies have not considered the relationships within gene-sets which may shorten its functionalities and applications. Here, we presented an analytical framework called Gene Set-based Local Hierarchical Clustering (GSLHC) approach which may provide biologically valuable insights on coordinated actions on functionalities and improved classification of heterogeneous subtypes on drug-driven responses. We successfully applied GSLHC on the Connectivity Map (C-Map) dataset with various gene sets from the Molecular Signatures Database (MSigDB). The GSLHC approach eliminated cell type effects that was obviously observed by IGA and showed significantly better performance than IGA on sample clustering and drug-target association. Furthermore, based on sets of significantly enriched gene sets, GSLHC identified 18 unknown compounds which functionally associated with the most correlated drug neighbors, that 8 of them contain putative anti-cancer activities. With extended applicability, GSLHC will facilitate the gaining of the biological insights on unknown drug discovery, drug repositioning, gene-set pattern diagnosis of common disease, and function-based class categorization of heterogeneous cancer subtypes.
