Computational Design of Protein-Ligand Interfaces - a Therapeutic Strategy

描述(申请人提供)：结合小分子的蛋白质可以通过隔离配体、刺激信号通路、将其他分子传递到作用部位以及作为体内诊断而起到治疗作用。尽管目前还不可能设计出可以与任何特定配体结合的蛋白质，但最近在从头开始酶设计方面的成功表明，这是可以实现的。目前的方法仍然不能预测最佳的氨基酸，甚至在配体周围的第一个壳中。具有部分共价性质的短程相互作用(如氢键、盐桥和阳离子相互作用)对于在结合位内实现精确定位通常是至关重要的，但由于它们的强度由相互作用官能团上轨道的几何形状、极性和极化率决定，因此很难建模。现有的对接技术很难处理灵活性 因此，没有考虑界面的结构塑性。W认为，蛋白质-配体界面的计算从头设计不仅可以扩大我们对分子识别所涉及的基本力的理解，而且如果能够克服某些技术限制，还可以促进蛋白质治疗学的发展。这项提议的目的是开发一种用于从头设计蛋白质-配体界面的计算方案。计算设计程序Rosetta将通过一个新的评分功能进行扩展，该功能使用基于知识的势来捕获蛋白质-配体界面的部分共价相互作用(PCI-KBP)。此外，还将实施一种基于片段的小分子构象采样方法。新的采样策略模拟了结合界面的配体灵活性，并利用了用于蛋白质设计的氨基酸旋转异构体采样的速度。计算模型的准确性将通过重新设计和对16个蛋白质突变体进行实验表征来评估，每个突变体都经过优化，可以结合16个相关化合物集中库中的一个小分子。将使用基于核磁共振(核磁共振)的筛选实验来确定整个16x16=256个组合的靶结合。核磁共振允许检测弱结合，确定结合亲和力，并在原子水平上详细验证结合部位。这种方法创建了设计界面的详细图，并通过对配体(衍生化)和蛋白质(突变)的化学修饰来捕捉对结合的影响。实验确定的结合亲和力矩阵将与Rosetta预测的矩阵进行比较，提供关于能量函数单个分量的准确性和采样策略效率的反馈。页码：1

DESCRIPTION (provided by applicant): Proteins that bind small molecules can act as therapeutics by sequestering ligands, stimulating signal- ing pathways, delivering other molecules to sites of action, and serving as in vivo diagnostics. Although the computational design of proteins that can bind to any given ligand is not yet possible, recent successes in de novo enzyme design suggests that it is within reach. Current methods still fail to predict optimal amino acids even in the first shell around the ligand. Short-range interactions with partial covalent character (e.g. hydrogen bonds, salt bridges, and cation-¿-interactions) are often critical for achieving precise positioning within the binding site but are difficult to model becaue their strength is determined by the geometry, polarity, and polarizability of orbitals attached to the interacting functional groups. Existing docking techniques have difficulty handling flexibility of the binding partners, so the structural plasticity of the interface is not taken into account. W believe that computational de novo design of protein-ligand interfaces can not only expand our under- standing of the basic forces involved in molecular recognition, but can also contribute to the development of protein therapeutics, if certain technological limitations can be overcome. The objective of this proposal is to develop a computational protocol for the de novo design of protein- ligand interfaces. The computational design program, ROSETTA, will be expanded through a new scoring func- tion that uses Knowledge-Based Potentials that capture Partial Covalent Interactions (PCI-KBP) at protein- ligand interfaces. Additionally, a fragment-based approach for sampling small molecule conformations will be implemented. The new sampling strategy models ligand flexibility at the binding interface and exploits the speed of amino acid rotamer sampling used for protein design. The accuracy of the computational models will be assessed through redesign and experimental characterization of a panel of 16 protein mutants, each optimized to bind one small molecule out of a focused library of 16 related compounds. Target binding will be determined for the entire set of 16x16=256 combinations using nuclear magnetic resonance (NMR)-based screening experiments. NMR allows detection of weak binding, determination of binding affinities, and verification of the binding site at atomic-level detail. This approach creates a detailed map of the designed interfaces and captures effects on binding through chemical modification of the ligand (derivatization) as well as the protein (mutation). The matrix of experimentally-determined binding affinities will be compared to those predicted by ROSETTA, providing feedback on the accuracy of individual components of the energy function and the efficiency of the sampling strategy. Page: 1