藉由粒線體相關疾病之基因型及表徵型的網路分析來預測疾病基因

Abstract

在遺傳疾病的研究上，基因與疾病之間的關係是我們感興趣的。其中我們更感興趣的是，是否仍有一些基因與特定疾病的關係被隱藏起來而未被發現或驗證。然而盲目的透過生物實驗的方法一一檢驗證其他的基因與疾病關連性，不僅曠日耗時，更需大量的金錢。根據文獻的回報，我們建立網路來連結各個基因與疾病。而從文獻裡，我們亦可估算出兩者之間的機率，藉以完成一個基因與疾病的網路，並建構演算法選出可能的致病基因。最後再利用交叉比對，決定模型的把關條件並且證明這個模型對於選取致病基因的可行性。

In the study of heritable diseases, we are interested in the relationship between genes and diseases. What we are more concerned about is if there are some hidden relationships which were not validated in literature reports. However, it costs time and money to clarify them one by one through biologic experiments. According to literature reports, we can not only build a network to connect genes and diseases, but also estimate the probability of this network. By this network, we can predict candidate genes which also cause diseases and are not observed. Finally, cross validation studies are carried out to decide thresholds of models and evaluate the performance of our methods proposed in the article. The results show that these new methods are promising.