COMPOUNDS FOR SELECTIVE KINASE INHIBITION

这个子项目是许多研究子项目中的一个 由NIH/NCRR资助的中心赠款提供的资源。子项目及 研究者（PI）可能从另一个NIH来源获得了主要资金， 因此可以在其他CRISP条目中表示。所列机构为 研究中心，而研究中心不一定是研究者的研究机构。 激酶抑制剂在各种激酶之间的交叉反应性是设计对某些激酶具有特异性的抑制剂分子的一大障碍。仅在最近，钌有机金属配体抑制剂分子被设计（在宾夕法尼亚大学），其对糖原合成酶激酶3（GSK-3）具有特异性，但支配这种特异性的原因还不清楚。一般来说，规则和分子机制，决定激酶抑制剂的特异性是一个相对未知的主题，仍然需要调查。更好地理解这个问题是重要的，因为激酶特异性抑制剂可用于破坏参与病理细胞中关键信号传导途径的激酶的活性，即，抑制与患病细胞相关的细胞信号传导途径可能具有潜在的治疗价值。在这里，我们开发了一个分层的计算策略，使用隐式和显式协议来表征抑制剂结合模式和亲和力（自由能）。我们的隐式方案是基于使用多个快照的激酶大分子从分子动力学（MD）模拟，允许不同的构象采样，因此能够捕捉蛋白质的灵活性。利用这种多构象对接策略，研究了小分子酪氨酸激酶抑制剂与这些激酶的不同结合模式。然后，一个更严格的明确的方法，涉及完全灵活的蛋白质和配体系统在明确的溶剂伞采样自由能计算将被用来完善铅结构的结合能。这一策略有望对我们正在研究的有机化合物抑制剂中激酶特异性的起源产生重要的影响。我们将比较钌基有机物抑制剂与三种激酶（即GSK-3、PIM-1和CDK-2）的结合特性。目前，我们已经能够进行10 ns的MD模拟（使用NAMD）的三个激酶（GSK-3，PIM-1，和CDK-2）在明确的水。我们还进行了模拟单框对接之间的基于辅酶的有机抑制剂分子与三个激酶，采用AutoDock。我们建议自动化的钌基有机抑制剂与三种激酶之间的模拟对接，通过我们的内部并行代码并行执行多构象对接。我们建议在teragrid上跨平台测试此并行代码，并为此请求20，000个SU。每个单帧对接需要相当于18个CPU小时的NCSA钨或SDSC数据星。对于每个激酶系统，我们将执行三个32处理器并行运行18小时，以处理从我们的MD模拟拍摄的100个快照。这相当于[18 CPU小时]*[32个处理器]*[每个激酶3次运行]*[三个激酶系统]=5800 SU。对于所得到的最低结合配置（两个激酶，即PIM-1和GSK-3），我们建议进行伞形采样模拟，以细化结合的结合自由能。这需要[24 CPU小时/处理器/ns]*[32个处理器]*[使用NAMD的1 ns MD/伞]*[7伞/激酶]*[2激酶]=10753 SU以获得自由能数据。我们要求大约4000个SU来测试我们的内部并行执行多构象对接。我们总共要求5800个SU +10753个SU +4000个SU = 20，554个SU，四舍五入为20，000个SU。我们在teragrid DAC资助下请求这些，因为这是我们第一次尝试运行跨平台模拟。(We我们的单平台并行应用程序的MRAC更新提案待定）。

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Cross-reactivity of kinase inhibitors amongst various kinases is a large obstacle in the design of inhibitor molecules that would possess specificity towards certain kinase enzymes. Only recently, ruthenium organometallic ligand inhibitor molecules were designed (at the University of Pennsylvania) that possess specificity towards glycogen synthase kinase 3 (GSK-3), but the reasons governing such specificity are not well understood. In general, the rules and molecular mechanisms that dictate kinase inhibitor specificity is a relatively uncharted subject that still requires investigation. A better understanding in this problem is important because kinase specific inhibitors can be used to disrupt the activity of kinases involved in crucial signaling pathways in pathological cells, i.e., inhibition of cellular signaling pathways associated with diseased cells could potentially have a therapeutic value. Here, we developed a hierarchical computational strategy using implicit and explicit protocol to characterize the inhibitor binding modes and affinities (free energies). Our implicit scheme is based on using multiple snapshots of the kinase macromolecule from a molecular dynamics (MD) simulation, allows sampling of different conformations, and therefore is able to capture the flexibility of the protein. The different binding modes of small molecule tyrosine kinase inhibitors with these kinases are examined using this multiple conformation docking strategy. Then a more rigorous explicit approach involving fully flexible protein and ligand systems in explicit solvent umbrella sampling free energy calculations will be used to refine the binding energetics of lead structures. This strategy is expected to throw significant insight on the origin of kinase specificity in the class of organometalic inhibitors we are studying. We will compare the binding characteristics of a Ruthenium based organometalic inhibitor to three kinases, namely GSK-3, PIM-1, and CDK-2. Currently, we have been able to conduct 10 ns MD simulations (using NAMD) of the three kinases (GSK-3, PIM-1, and CDK-2) in explicit water. We have also performed simulated single frame docking between the ruthenium-based organometalic inhibitor molecules with the three kinases by employing AutoDock. We propose to automate the simulated docking between the ruthenium based organometalic inhibitor with the three kinases to perform the multiple conformation docking in parallel through our in-house parallel code. We propose to test this parallel code across platforms on the teragrid and request 20,000 SUs for this purpose. Each single frame docking requires an equivalent of 18 CPU hrs on NCSAs tungsten or on SDSCs datastar. For each kinase system we will perform three 32-processor parallel runs for 18 hrs to process 100 snapshots taken from our MD simulations. This amounts to [18 CPU hrs]*[32 processors]*[3 runs per kinase]*[three kinase systems]=5800 SUs. For the resulting lowest binding configurations (for two kinases, namely PIM-1 and GSK-3), we propose to perform umbrella sampling simulations to refine the binding free energies of binding. This requires [24 CPU hrs per processor per ns]*[32 processors]*[1 ns MD per umbrella using NAMD]*[7 umbrellas per kinase]*[2 kinases]=10753 SUs to obtain the free energy data. We request about 4000 SUs to test our in-house parallel for performing the multiple conformation docking. In total we request 5800 SUs+10753 SUs+4000 SUs=20,554 SUs rounded off to 20,000 SUs. We request these under a teragrid DAC grant because this is our first attempt to run cross platform simulations. (We have an MRAC renewal proposal pending for our single platform parallel applications).