蛋白質與小分子之強固評分函數的開發與應用

Abstract

蛋白質與小分子之間作用力的評分函數，在電腦輔助藥物設計，在先導藥物的虛擬篩選中，以及在預測化學小分子作用標的裡扮演重要的角色。本論文我們建立了一組強固評分函數和應用它在計算藥物設計領域。一般的最小平方法迴歸分析（OLS）已經被大量運用在建立蛋白質與小分子之間的評分函數上。然而OLS對於離群值(outliers)是很敏感的，所以使用OLS建立出來的模型很容易受到離群值或訓練資料選擇不同的影響。另一方面，原子電荷的決定也被認為是很重要的，因為靜電力的作用被認為是生物分子間結合的一個關鍵因素。我們提出新的評分函數是基於AutoDock4評分函數的形式，並且利用量子力學與分子力學得到更嚴格的原子電荷。除此之外，我們更建立了一套使用強固迴歸分析來訓練強固評分函數的方法。換句話說，訓練資料之中離群值的問題可以被解決。我們新的評分函數在評定測試中，表現得比大部分其他評分函數好，這包含結合親和力的預測和的能夠鑑別正確結合位置與構形的能力。 在本論文第一章，我們將探討現存不同類別的評分函數，並且討論它們可能的局限和適合的應用範圍。第二章將簡單介紹分子嵌合軟體AutoDock，它是本研究的起點。第三、第四章節將分別闡述強固評分函數的建立過程和它的效能測試。在第五章，這是一個新穎的網站應用，名叫idTarget，目的是探索與鑑別可能的蛋白質標的物。

The scoring functions for protein-ligand interactions plays central roles in computational drug design, virtual screening of chemical libraries for new lead identification, and prediction of possible binding targets of small chemical molecules. We have developed the robust scoring functions and applied it for computational drug design. Ordinary least-squares (OLS) regression has been used widely for constructing the scoring functions for protein-ligand interactions. However, OLS is very sensitive to the existence of outliers, and models constructed using that are easily affected by the outliers or even the choice of the data set. On the other hand, determination of atomic charges is regarded as of central importance, because the electrostatic interaction is known to be a key contributing factor for biomolecular association. Our new scoring functions were based on the functional form of the AutoDock4 scoring function and using more rigorous charge models derived from quantum mechanics and molecular mechanics. On top of that, we developed a protocol for calibrating the robust scoring function by using the robust regression analysis. In another word, the problem of outliers in the training set can be solved. The assessments show that our new scoring functions outperformed most of other scoring functions on predicting binding affinity and discriminating the native pose from decoys. In the first chapter of the present dissertation, we will explore the foundations of different classes of scoring functions, their possible limitations, and their suitable application domains. The second chapter introduces the docking program AutoDock which is the basis of this study. In Chapter 3 and 4, the development of the robust scoring functions and its assessments will be described, respectively. In Chapter 5, a novel application of web service, namely idTarget, which aims to identify protein targets, will be presented.