找出導致腫瘤發生的生化路徑對於癌症診斷或是治療藥物的研發製造上都是極為重要的工作。雖然目前已有多種生物資訊方法已被使用在找尋癌症病人特有的基因表現路徑或是特徵生化路徑，但是由於癌症在人體上呈現的變異性，因此對於癌症生化路徑仍然無法準確地預測。在此次計畫中，我們設計了一個新演算法可以有效率的預測癌症的特徵生化路徑，而且將這個方法用於口腔癌的研究上。利用這個方法，我們已找到與促使腫瘤持續生長相關的生化路徑，而且其中一些路徑已被證明的確在口腔癌上具有重要調控的機制。

Identification of oncogenic pathways is critical in cancer diagnosis and drug design. Recent, a number of bioinformatics methods have used to discovery dysregulated pathways or subnetwork biomarkers of cancers. However, the accurate rate of causal pathways predictions is still need to be improved in cancers with a high degree of heterogeneity. In this study, we designed an effective algorithm for predicting subnetwork marker. The algorithm is used search the subnetwork marker of the maximum relative variation compared with cancer and normal samples. We applied the algorithm to identify significant disordered pathways in oral squamous cell carcinoma and obtained expectable results. The subnetwork found contains crucial pathways of oral cancer progression and is consisted with previous studies.