橢圓排序導引階層分群樹(HCT-R2E)應用在基因表現資料的矩陣視覺 化及群集分析上，是一種很有效的方法。它可以同時對基因表現資料提供較一致的局部群集和較佳的全域群組狀態。然而和傳統的數理式的群集分析一樣，橢圓排序導引階層分群樹方法僅利用到基因微陣列表現資料卻未考慮到把這些已知基因功能的屬性結合到分群演算裡。在本研究中，我們結合微陣列資料之基因所代表的生物知識，計算一個新的距離尺度，當作橢圓排序導引階層分群樹法使用的距離尺度。新的距離尺度的採用可以同時獲得群集後基因表現的相似性與基因功能屬性的同一性。以結合生物知識為基礎的橢圓排序導引階層分群樹法應用在酵母菌細胞週期和老鼠腦細胞這兩種微陣列資料，我們發現結果不僅保存原本橢圓排序導引階層分群樹法所具有的分群排序性質，也同時提供更相關及有意義的生物註解資訊去幫助識別基因的功 能屬性。 The hierarchical clustering tree (HCT) guided by a rank-two ellipse seriation (R2E) is an effective method to identify coherent local clusters and better global grouping patterns simultaneously in gene expression profiles. Like most other mathematical clustering methods, the HCT-R2E conducted only on the statistical characteristics of gene expression data while the known gene functions was not considered in the clustering process. In this study, we incorporate these information to create a new distance metric for HCT-R2E. The new distance metric captures both expression pattern similarities and biological function agreements. With cases studies on the microarray data of the yeast cell-cycle and mouse mesencephalon data. we shown the biological knowledge-based HCT-R2E not only preserves the desirable properties of its own its own but also identifies genes that are more relevant and meaningful to biological annotations.
