化合物抑制劑與蛋白激酶結合之嵌合計算研究:胺基酸支鏈可動性及活化中心氫鍵限制的影響

Abstract

我們探討兩種蛋白激酶，CDK2和c-MET的分子嵌合計算。此兩種蛋白激酶在藥物發展上是許多實驗學家所感興趣的。 在CDK2蛋白激酶的部分，我們研究在虛擬篩選時，是否擺動胺基酸支鏈加上活化中心氫鍵限制可以提高enrichment factor(EF)。我們使用Directory of Useful Decoys (DUD)資料庫作為測試基準資料庫。平均來說，將活化中心氫鍵限制可以使EF提高2倍；若單獨擺動胺基酸支鏈僅能使EF微幅增加。兩個效應同時在嵌合計算開啟所得到的EFs和個別效應所得到的EFs並沒有明顯的增加。有趣的是，某些胺基酸支鏈相較其他而言會對EF有很大的影響。這些計算結果對於虛擬篩選是有幫助的，在大型的分子數據資料庫以獲得更好的命中。 除了CDK2激酶外，我們也使用相同的分子嵌合計算程序對c-MET激酶做研究。c-MET為一種蛋白質激酶，在多細胞生物中扮演重要的角色，如調控細胞的增殖、移動、侵襲、轉移與血管新生等，且對於胚胎發育和傷口癒合是不可或缺的；然而，c-Met的過量表現或突變，也是造成人類癌症的原因之一。 首先，我們從蛋白質資料庫可得的結晶構型，挑選10個c-MET抑制劑複合物做分子嵌合計算，研究其再現性。第二，我們再選擇5個c-MET抑制劑複合物做交叉分子嵌合，藉此研究當c-MET在不同的結晶構型時，如何再現小分子的構型。第三，我們選擇前3名的c-MET結晶構型進行虛擬篩選，計算10個已知具有抑制力的小分子，在嵌合計算中可篩選出多少個活性分子。同時也設定擺動胺基酸支鏈及活性中心氫鍵限制。我們發現，當設定3個可動胺基酸支鏈加上可形成氫鍵原子做限制時，和其他組合相較之下結果較佳。最後，我們藉由以上較佳的條件選擇和設定，高速虛擬篩選40萬個化合物，並分析數個具有較佳親和力的化合物與c-MET之間的作用力和結合圖形，而這些計算結果將有助於實驗學家設計與搜尋c-MET抑制劑。

關鍵字:激酶、分子嵌合計算、虛擬篩選In the present study, we carried out docking computation of compounds against two kinases, CDK2 and c-MET, using GOLD program. Both are of pharmaceutical interest. In the part of CDK2 kinase, we investigated if allowing side chains to move and applying hydrogen bond constrains can enhance enrichment factor(EF) in virtual screening. To this end, compounds from DUD database were used for benchmark. The computations gave that applying hydrogen bond constraints enhance EF, on average, by about a factor of 2, and allowing side chains to move only enhance EF slightly. With both effects turned on in docking computation, the calculated EFs do not enhance significantly compared with EFs obtained from individual effect. Interestingly, some side chains have more significant enrichment effects than others. These computed results should be useful for virtual screening over large compounds databases against CDK2 kinase in order to obtain better hits. In addition to CDK2, we also investigated c-MET kinase using similar docking procedure. c-MET (MET) is a kinase protein that plays an important role in multi-cells, including regulation of proliferation, motility, invasion, migration, and angiogenesis. It is also essential for embryonic development and tissue damage repair. However, overexpression of c-MET or mutations occurs in many human cancers. First, we carried out docking computations for 10 c-met inhibitor complexes with crystal structures available in the protein data bank in orderto examine their reproducibility.Second, we choose 5 complexes to do crossing docking and investigate how ligand conformations can be regained when protein structures from different complexes were used. Third, we selected top3 protein structures from these 5 complexes for subsequent virtual screening. For benchmark, we constructed a group of 1000 compounds consisting 10 active and 990 decoy compounds, and see if 10 active compounds can be screened out in docking computation.Effects of side chain flexibility and hinge hydrogen bond constraints were examined as well. We found that the results obtained byallowing 3 residues to move and constraining2hinge hydrogen bonds simultaneouslyare better than results obtained by other combinations. Finally, virtual screening for 400,000 compounds of c-MET was carried out. The interactions between top-ranked compounds and c-met were analyzed and discussed. These computed results and analysis should be of aid in design and discovery of c-met inhibitors.

key words:kinase、docking computations、Virtual Screening