基于蛋白结构的成药性预测新方法研究

Poor drugability (ADME and toxicity) is indeed an important reason for stopping further development of drug candidates. The experimental evaluations of drugability are time-consuming and expensive, and therefore there is a tremendous need for developing high quality and reliable in silico models to evaluate drugability for compounds in libraries and molecules virtually synthesized with computers. In this proposal, we will focus on the development of new theoretical approaches for drugability predictions based on the characterization of protein-ligand interactions for three important ADMET-related proteins, including P-glycoprotein, hERG and PXR. First, a new MM/GBSA protocol based on variable dielectric GB model and a fast approach to estimate the conformational entropy based upon atom-weighted solvent surface area calculations will be developed, and then the best strategy for the accurate estimation of binding free energies will be determined. Then, three protein-based approaches for drugability predictions will be developed and compared, which include the approach based on molecular docking, molecular dynamics (MD) simulations and free energy calculations, the approach based on protein-ligand free energy decomposition and machine learning techniques, and the approach by combining structure-based pharmacophore modeling and machining learning techniques. We believe that our research will be very promising for the development of new theoretical methods to estimate binding free energies, the understanding of the recognition mechanism between ADMET proteins and drug molecules, and the accurate estimation for important drugability descriptors.

药代动力学或毒性（ADMET）等成药性缺陷是导致药物开发失败的主要原因。成药性实验评估代价高，周期长，因此发展可靠的成药性预测模型具有非常重要的研究意义和重大的应用前景。本课题拟针对3个重要的ADMET相关蛋白（P糖蛋白、hERG钾离子通道及孕烷X受体），发展基于蛋白-配体相互作用分析的成药性预测方法，主要研究内容包括：(1). 发展基于可变介电常数的MM/GBSA方法，完善基于加权溶剂可及表面的高效构象熵计算方法，并探索对ADMET蛋白最佳的结合自由能计算策略；(2). 发展并比较3种基于蛋白结构的成药性预测方法，即基于分子对接、分子动力学模拟和自由能计算的预测方法，基于蛋白-配体结合自由能分解和机器学习的预测方法，以及基于蛋白药效团模型和机器学习的预测方法。本项目对于揭示ADMET相关蛋白和药物分子之间的相互识别机制以及关键成药性参数的可靠预测都具有重要意义。

成药性缺陷是导致候选药物开发失败的主要原因之一，并可能在治疗过程中引起多种严重后果。为了有效提高药物开发的效率和成功率，需要在药物研发的早期阶段对化合物的成药性进行评价并尽早排除成药性差的化合物。成药性评价的实验评价昂贵、周期长、工作强度大，因此发展高效可靠的成药性理论预测模型具有重要意义。本项目针对多个重要的ADMET相关蛋白，发展基于蛋白-配体相互作用分析的成药性预测方法，主要研究内容包括：(1). 发展基于可变介电常数的MM/GBSA方法，并探索对ADMET蛋白最佳的结合自由能计算策略；(2). 发展并比较多种基于蛋白结构的成药性预测方法，即基于分子对接、分子动力学模拟和自由能计算的预测方法，基于蛋白-配体结合自由能分解和机器学习的预测方法，以及基于蛋白药效团模型和机器学习的预测方法。在本项目的资助下，共发表SCI研究论文23篇。本项目对于关键成药性参数的可靠预测具有重要意义。