【作者】 李林；

【导师】 肖奕；

【作者基本信息】 华中科技大学 ， 理论物理， 2011， 博士

【摘要】 大部分蛋白质需要通过与其他蛋白质或分子的相互作用来完成其生命功能,对蛋白质-蛋白质相互作用的研究需要蛋白质复合体的三维结构。与测量蛋白质单体结构相比,实验上测定复合体三维结构相对比较困难。因此,目前实验测定的蛋白质单体的结构数量已经远远大于复合体的结构数量。通过蛋白质-蛋白质对接方法可以帮助预测蛋白质复合体的三维结构。在很多蛋白质-蛋白质对接方法中,几何匹配是最为重要的打分方式,因此如果能够改进目前的几何匹配模型,将会对很多对接方法都有所帮助。正确的生物学信息可以帮助预测复合体三维结构,错误的生物学信息会导致预测失败,所以不可靠的生物学信息难以被直接利用。许多蛋白质分子对接方法都是基于快速傅里叶变换(Fast Fourier Transform, FFT)算法。基于FFT方法的对接算法都以几何匹配打分为主要的打分依据。我们在传统的FFT对接方法的基础上,开发了基于原子溶剂化参数(Atomic Solvation Parameters, ASP)模型的新的对接方法——ASPDock。ASPDock用ASP打分替代了FFT方法中原有的几何匹配打分。FFT方法中的ASP打分在计算方法上与纯几何匹配有一些相似,但比几何匹配有着更加明确的物理意义。纯几何匹配只是反映了蛋白质复合体的几何性质,而ASP模型能够成功地计算蛋白质-蛋白质复合物的结合自由能。我们使用ASPDock计算了21组实验已经测定了结合能的复合体。当格点大小取值为1 ?时,得到的21组复合物的ASP打分与实验测定的结合能之间的相关性(r≈0.69)要高于几何匹配与实验结果的相关性(r≈0.48)。这个结果表明,ASPDock的打分方式要比几何匹配的打分方式更为可靠。在针对Benchmark3.0的124组复合体的测试中,ASPDock的预测成功率也明显优于几何匹配。在蛋白质对接的算法中,利用可靠的生物学信息可以有效地提高预测的成功率。因此很多小组开发了各种不同的理论方法来预测蛋白质的相互作用位点信息。然而,目前各个小组的理论预测方法并不是完全可靠的。为了能够利用能够这些不完全可靠的理论预测方法来帮助对接算法,我们开发了一个软限制(softly restricting method,SRM)的方法来利用这些信息。SRM方法在ASPDock的基础上,通过提高某些关键残基的权重的方式来使采样倾向于在集中那些关键残基附近。SRM方法对正确的信息非常敏感,而对错误的信息不太敏感,这样可以有效地降低使用不可靠信息的风险性。我们在99组蛋白质复合物上进行了测试,结果表明,即使利用有风险的理论预测位点信息,SRM方法依然可以提高预测的成功率和增加预测到的近天然态结构的数目。ASPDock和SRM方法也在CAPRI(Critical Assessments of PRediction of Interactions)的第18轮和第19轮比赛中取得了成功。通过ASPDock和SRM方法的结合使用,我们预测得到了CAPRI T40和T41的高精度结构,LRMSD分别为2.35 ?和1.41 ?.更多还原

【Abstract】 Most proteins interact with other proteins or molecules to perform their biological functions. The details of protein-protein interactions need 3D structures of complexes. However, it is relatively difficult to measure the structures of protein complexes experimentally and thus the number of available complex structures is still limited in comparison with monomer protein structures. Therefore, it is helpful to use computational approaches to predict structures of protein complexes. Shape complementary is the most important scoring function for many protein-protein doking algorithms, thus it will be helpful for many doking algorithms if the shape complementary model could be improved. Correct biological information helps to predicted structures of complexes, however, incorrect information will lead to a failure for predition, thus the unreliable biological information is difficult to utilize.We have developed a new Fast Fourier Transform (FFT) docking algorithm:ASPDock,which is based on Atomic Solvation Parameters (ASP) model instead of pure shape complementary. ASP model in FFT method is similar to shape complementary in practical calculation but illustrates the binding mechanism from a clearer view of physics. Shape complementary, which plays the most important role in almost all the docking algorithms, mainly focuses on the geometric feature of protein complexes. ASP model has made great success in calculations of binding free energy of proteins. We used ASPDock to calculate ASP scores of protein complexes and test it on 21 complexes with the experimentally determined binding free energies. The results show that the ASP score has a much stronger correlation (r≈0.69) with binding energy than shape complementary score (r≈0.48) on 1.0? grid space. This test indicates that, comparing with pure shape complementary, the ASP model is more reliable for docking. ASPDock has also been tested on a large dataset, benchmark3.0, which contain 124 complexes and it also performs better than that based on pure shape complementary.Reliable biological information helps to predict protein-protein complex structures. Several methods have been proposed to predict protein binding site theoretically. However, theoretical predicted binding sites is usually unreliable. We developed a softly restricting method (SRM) to use the unreliable binding site information. Based on ASPDock, SRM enhances the weight of the expected binding site residues and softly constrain receptor and ligand to bind around the expected binding site. This SRM algorithm is proved to be sensitive to correct binding information and insensitive to wrong binding site information, which should decrease the risk of theoretical binding site information. We performed SRM on 99 complexes. It shows that using theoretical predicted binding sites information, SRM enhanced the hit count number and success rate.ASPDock and SRM have been proved to be successful in CAPRI round 18 and 19. Combining ASP docking method and softly restricting method, we got high quality hits for CAPRI T40 and T41, and the best LRMSD are 2.35 and 1.41?, respectively.更多还原