Receptor ligand binding: complex structure prediction and affinity evaluation

小分子与蛋白质活性位点的结合自由能的可靠和快速计算对于基于结构的药物设计和理解蛋白质与配体的相互作用至关重要。严格的治疗涉及计算昂贵的方法，如自由能摄动方法，然而，这与药物筛选或铅发现等快速方法的需求不相容。反过来，常用的快速方法涉及经验推导的评分函数，通常不包括受体和/或配体的灵活性。因此，这种方法在准确性上是有限的。该项目旨在基于可用的结构数据，结合用于计算自由能差的物理有效能量函数，开发一种快速的蛋白质配体构象搜索。第一步涉及基于几何考虑的蛋白质配体构象的有效预测，旨在有效地采样配体-受体构象的可用构型空间。随后，熵估计用于加权的抽样构象，以产生规范集成。这些集合可以通过（集合）平均基于物理化学（力场）的能量函数和有效的连续静电方法来评估溶剂化自由能来精确确定结合能。这两个步骤——生成替代构象和使用正则系综平均计算自由能——被期望能够准确有效地估计相对结合亲和度。在随后的步骤中，该方法可以扩展到新的先导物发现或对接策略，充分考虑到蛋白质和配体的灵活性。

The reliable and fast computation of binding free energies of small molecules to the active sites of proteins is crucial for structure-based drug design and for the understanding of protein-ligand interactions. Rigorous treatments involve computationally expensive methods like free energy perturbation approaches, which are, however, incompatible with the need for fast methods in e.g. drug screening or lead discovery. In turn, commonly used fast methods involve empirically derived scoring functions and usually do not include receptor and/or ligand flexibility. Hence, such methods are inherently limited in accuracy. The proposed project aims at the development of a fast conformational search for possible proteinligand conformations based on available structural data combined with a physical effective energy function for the computation of free energy differences. The first step involves the efficient prediction of protein-ligand conformations based on geometric considerations, aimed at efficiently sampling the available configurational space of ligand-receptor conformations. Subsequently, entropy estimates are used for weighting of the sampled conformations, in order to generate canonical ensembles. These ensembles allow an accurate determination of binding energies by (ensemble) averaging an energy function which is based on physical chemistry (force field) and an efficient continuum electrostatic approach to evaluate solvation free energy. Together, these two steps — the generation of alternative conformations and the computation of free energies using canonical ensemble averaging—are anticipated to allow an accurate and efficient estimation of relative binding affinities. In subsequent steps the method may be extended towards novel lead discovery or docking strategies, fully taking into account protein and ligand flexibilities.