Computational Methods for Modeling Drug Kinetics

小分子药物仍然是现代医学的重要组成部分，尽管传统的药物开发模式面临着成功率下降和成本上升的问题。传统药物化学的目的是鉴定具有高亲和力的结合蛋白靶点的药物。在此基础上，计算化学家开发了通过模拟药物-蛋白质分子间相互作用来估计药物-蛋白质结合亲和力的方法。在过去的10年里，化学家们发现，药物与靶标结合和分离的速度可以更准确地预测药物的疗效。特别是，保持结合时间较长的药物（即延长停留时间，tD）可能具有高疗效。长效药物不太可能被取代、代谢或干扰另一种细胞成分，这可以提高疗效，减少不良反应。动力学效应是最显著的药物结合在蛋白质的隐性位点。从这些位点解离需要药物通过一组复杂的动态路径连接中间结合状态。这些途径可能涉及比简单药物扩散更长的时间尺度的过程，如药物和蛋白质的构象变化、脱溶和特定分子间接触的破坏/建立。对于共价调节剂药物，解离也包括破坏共价键。药物解离路径的模拟将提供缺失的机制数据，这将使长期药物的合理设计成为可能。该计划的长期目标是通过计算预测药物结合动力学来设计更有效的药物，这将通过我们的短期目标来实现：(A)药物-蛋白质结合的马尔可夫状态模型（mms）， (B)模拟共价修饰剂药物的方法，(C)药物-蛋白质分子间相互作用的模型，以及(D)通过计算激酶抑制剂的结合动力学来验证和应用这些方法。药物结合动力学将通过将一组中间结合模式描述为通过动态过渡路径连接的MSM状态来建模。元动力学将用于识别中间状态，然后使用路径采样方法来计算状态之间的转换速率。将开发新的模型来更真实地描述药物-蛋白质相互作用的排斥和分散成分。这些方法将使有效和准确地计算结合动力学成为可能。在验证阶段，我们将评估我们的方法预测药物抑制激酶蛋白的动力学。许多激酶都是药物靶点，但它们的构象灵活性和激酶家族内的高度同源性给设计选择性药物带来了困难。我们将使用激酶-药物相互作用的实验数据验证我们的方法，然后将这些方法应用于设计长td激酶靶向药物。

Small-molecule drugs remain a critical component of modern medicine, although the traditional drug development model has faced declining success rates and escalating costs. Traditional medicinal chemistry has aimed to identify drugs that bind to their protein targets with high affinity. In aid of this, computational chemists have developed methods to estimate drug-protein binding affinities by modeling the drug-protein intermolecular interactions. Over the last 10 years, chemists have found that the rates at which a drug binds and dissociates from its target can be more predictive of a drug's efficacy. In particular, drugs that remain bound for longer periods (a.k.a. prolonged residence time, tD) can have high efficacy. Long-tD drugs are less likely to be displaced, metabolized, or interfere with another cellular component, which can yield improved efficacy with fewer adverse effects. Kinetic effects are most significant for drugs that bind in recessed sites of proteins. Dissociation from these sites requires the drug to pass through a complex set of dynamical paths connecting intermediate binding states. These paths can involve processes that occur on much longer timescales than simple drug diffusion, such as conformational changes of the drug and protein, desolvation, and breaking/making of specific intermolecular contacts. For covalent-modifier drugs, dissociation also involves breaking covalent bonds. Simulation of drug dissociation paths will provide the missing mechanistic data that will enable rational design of long-tD drugs. The long-term objective of this program is to design more effective drugs by predicting their binding kinetics computationally, which will be achieved by our short-term objectives: (A) Markov state models (MSMs) of drug-protein binding, (B) methods to model covalent-modifier drugs, (C) models for drug-protein intermolecular interactions, and (D) validating and applying these methods by calculating the binding kinetics of kinase inhibitors. The drug binding kinetics will be modeled by describing the set of intermediate binding modes as the states of a MSM that are connected by dynamical transition paths. Metadynamics will be used to identify the intermediate states, then path sampling methods will be used to calculate the rates of transitions between states. New models will be developed to describe the repulsive and dispersion components of the drug-protein interaction more realistically. These methods will make it possible to calculate binding kinetics efficiently and accurately. In the validation phase, we will assess our methods for the prediction the kinetics of drugs inhibiting kinase proteins. Many kinases are drug targets, but their conformational flexibility and the high homology within the kinase family has made it difficult to design selective drugs. We will validate our methods using experimental data on kinase-drug interactions then apply these methods to design long-tD kinase-targeting drugs.