發展基於分子動力模擬能量分解的新穎自由能計算方法

Abstract

能夠準確預測小分子與生物分子的結合在電腦輔助藥物設計上扮演了相當重要的角色，因為這些方法能加速先導化合物的產生與優化。這些方法在現在大約可分為兩大類，包含了分子嵌合與評分與自由能之方法。自由能方法需要構形取樣，反之，嵌合與評分的方式往往都只聚焦在單一的構型上面。也因此自由能方法需要較多的電腦計算時間與資源。 我們認為在預測蛋白質與配位體的結合能力仍然需要考慮到動態的影響，因為在結合的過程當中蛋白質與配位都是可運動的狀態。因此我們使用了linear interaction energy方法做為起點，因為這個方法是自由能方法中的一種而且只需要兩個分子模擬就能得到結合的能量。這個方法跟其他一樣屬於自由能方法的FEP與MM-PBSA來的更為經濟。 在我們的研究當中，利用GROMACS來做分子模擬且使用GROMACS的參數。預測結合的自由能我們除了使用最原始的LIE方法的公式外，還加入了配位體與配位體之間的能量項進去試著去得到更好的預測結果。另外，我們也使用了許多的評分的函數來與我們的結果做比較。

Accurate methods for predicting the affinity of a small molecule with a protein or other biomolecule play a crucial role in computational drug design because these predictions can speed the lead generation or lead optimization. Nowadays these methods can be categorized as docking and scoring and free energy method. Free energy method, in contrast with docking and scoring, which focus on a single bound conformation use conformation sampling to generate thermodynamic averages. Therefore, free energy methods are required more computer time than docking and scoring approaches. We consider that the dynamic effect of the protein-ligand complex should be still included in evaluating binding affinities because in the binding process, protein and ligand are flexible. Therefore, our study started from the linear interaction energy (LIE) method, which is one free energy method and only required two simulations, complex and ligad only. Besides that, it is also more economical than other free energy methods, such as the FEP and the MM-PBSA method. In our study, we used GROMACS to perform simulations with GROMACS force field parameters. For predicting the free energy of binding, we not only used the original LIE equation, but also we added ligand-ligand interaction into the evaluating scheme trying to obtain a better prediction model. In addition, we used several scoring functions, ChemScore, DrugScore, XScore, etc, to compare with our results.