利用支援向量機器預測特定位置突變所引起的蛋白質穩定性改變

Abstract

預測特定位置突變所引起的蛋白質穩定性改變是生物學上的重要議題，將序列及結構資訊有效的轉換為能量參數，將有助於蛋白質穩定性及功能的分析。近年來許多團隊將心力投注於單點特定位置突變及整體穩定性之實驗數據之間關連性。在這篇論文我們利用支持向量機器預測特定位置的單點突變所引起的蛋白質熱穩定性改變。基於前人的理論基礎，我們以八種不同的編碼方式將結構或序列資訊轉換為特徵向量，用以測試三個經由特定條件由線上資料庫ProTherm得出之公開資料集S1615, S2048以及S1396，並以預測平均準確率及Matthews相關係數評量我們的實驗成果。實驗數據顯示我們的方法可與目前最好的方法相當，並進一步能單以一級序列資訊改進在相同條件下預測準確率及相關性，這在醫療科技及蛋白質工業中缺乏次級以上結構資訊的多數情況下有實用性價值，利用本方法以電腦計算並預測特定位置突變所引起的蛋白質穩定性改變，我們可大幅減低傳統實驗的時間及成本。

To predict the effect of site-specific mutation on protein stability and function has been an important issue of protein science. Turning Sequence and structure information into energetic parameters enables us to predict and analyze protein function. In this experiment we predict the thermostability of single point mutation using the support vector machine. Based on previous knowledge of thermostability prediction of single point mutation, we use eight different encodings to transform sequence information and structure information into feature vectors to test the three public datasets extracted by certain filters from ProTherm. We use average accuracy and Matthews correlation coefficient of prediction of thermostability to evaluate our experiment results. The results show that our methods are comparable with the best current methods. Further more, we can predict the thermostability of single point mutation using the support vector machine by sequence information only when further information is not available yet.