- 學位論文
利用微陣列擾動數據及蛋白質功能資料 建置MAPK生物路徑預測系統
MAPK Biological Pathway Prediction System Based on Microarray Perturbation Data and Protein Function Information
摘要
MAPK 訊號傳遞路徑的機制廣佈於真核生物細胞中，且受外在環境變化而調控其轉錄作用。本研究是利用酵母菌基因之微陣列晶片實驗資料，來預測 MAPK 訊號傳遞路徑組成順序。我們利用了兩組實驗數據，一組包含了與訊號傳遞路徑有關的56種實驗，另一組則包含了300種不同突變與化學處理。計算方法是計算預測的路徑中基因(mRNA)與鄰近的基因(mRNA)表達量的皮爾森相關係數絕對值總和分數，然後將分數做排序後，最高分的路徑為預測之 MAPK 訊號傳遞路徑，從預測結果得知，真實 MAPK 路徑會落在所有可能路徑排名前15%中。 然後，我們再加入 蛋白質功能與功能 (FFC)的資料，來提昇路徑預測的排名。由預測結果得知，在加入 FFC 資料後，真實 MAPK 路徑會並無明顯增加，只有在Hypotonic中有將排名提升。另外，我們還有預測各路徑間的功能相似情形以及將我們的方法應用在蛋白質複合體的系統，預測組成蛋白質複合體的各個子單元順序。 我們將目前所作的資料建置成網頁，且提供下列功能:(i) 預測最有可能的MAPK 訊號傳遞路徑，按皮爾森相關係數絕對值總和後的高低排名，(ii)基因與基因間兩兩之Pearson相關的值，(iii)預測蛋白質複合體順序路徑，(iv)提供兩支Delphi介面程式，用來計算蛋白質功能與功能相關性機率值。
並列摘要
MAPK signal transduction pathways are widespread in eukaryotic cells, and changes in the external environment could activate these pathways. This study makes use of yeast gene microarray experimental data to predict the MAPK signal transduction pathway order. Two sets of experimental data are used, a group containing the signal transduction pathway data of 56 experiments, the other group includes 300 different mutations and chemical treatment experiments. The calculation is based on computing the absolute sum of Pearson correlation coefficient (PCC) between a mRNA and the nearby mRNA expression. The total score of a pathway is given by the sum of the individual pair along the pathway. Pathway has the maximum score is the predicted pathway. It is found that the real MAPK pathway fall on the first 15% among all possible pathways. Furthermore, we integrate the Function-Function Correlation (FFC) data, to improve the pathway forecast ranking. After joining the FFC data a few results get improved slightly. Finally, we use this method in the protein complex systems, forecasting the sub-units order in assembling a protein complex. A web based service is set up which offer the following features: (i) forecast the most likely MAPK signal transduction pathway for yeast, according to microarray data, (ii) compute the PCC for a pair of mRNAs, (iii) predicting the assembling order for a protein complex, and (iv) two Delphi programs for computing the FFC values for a pair of proteins.