Virtual Drug Screening in the Chemical Space Accessible by Chemical Synthesis

现代药物开发方法从确定靶标和虚拟筛选“类药物”有机分子开始。这需要两个相互交织的挑战：首先，“配体”必须是功能性的，其次，它必须是可合成的。通常，由于竞争专利，以前研究过的物质类别是要避免的。在这里，我们提出了一种虚拟药物筛选的计算方法，该方法结合了用于分子功能评估的机器学习现代技术，基于规则的化学反应建模以保持合成合理性，以及一种受进化算法启发的优化方法。交叉和突变操作在基于规则的框架中进行，旨在将搜索限制在化学上真实的分子上。基于最近对表征重要性的观察，我们提出，为反映化学家对结构式和基本反应的直觉而构建的图语法构成了化学空间遍历和学习功能重要特征的自然表征框架。我们的目标是开发一个全面的框架和附带的软件，用于虚拟药物筛选，将对申请人已经可用的一套筛选进行测试。作为提案的一部分，我们还将为这种虚拟筛选方法设计基准程序，包括与莱比锡大学的合作伙伴进行实验验证。

Modern approaches to drug development start with the identification of a target and virtual screening of “drug-like” organic molecules. This entails two intertwined challenges: first the “ligand” must be functional and second it must be synthetically accessible. Often, previously studied substance classes are to be avoided due competing patents. Here we propose a computational approach to virtual drug screening that combines modern techniques of machine learning for functional evaluation of molecules, rule-based modelling to chemical reactions to maintain synthetic plausibility, and an Evolutionary Algorithm-inspired optimization approach. Cross-over and mutation operations are phrased in the rule-based framework, aiming at restricting the search to chemically realistic molecules. We propose, based on recent observations on the importance of representation, that the graph grammars built to reflect the chemist’s intuition of structural formulae and elementary reactions constitute a natural representative framework for both the traversal of chemical space and for learning functionally important features. We aim at the development of a comprehensive framework and accompanying software for virtual drug screening that will be tested against a suite of screening that are already available to the applicants. A part of the proposal we will also devise benchmarking procedures for this kind of virtual screening approaches including experimental validation with collaboration partners at University of Leipzig.