Wright Trudi, Gillet Valerie J, Green Darren V S, Pickett Stephen D
Krebs Institute for Biomolecular Research and Department of Information Studies, University of Sheffield, Western Bank, Sheffield S10 2TN, United Kingdom.
J Chem Inf Comput Sci. 2003 Mar-Apr;43(2):381-90. doi: 10.1021/ci0255836.
This paper addresses a major issue in library design, namely how to efficiently optimize the library size (number of products) and configuration (number of reagents at each position) simultaneously with other properties such as diversity, cost, and drug-like physicochemical property profiles. These objectives are often in competition, for example, minimizing the number of reactants while simultaneously maximizing diversity, and thus present difficulties for traditional optimization methods such as genetic algorithms and simulated annealing. Here, a multiobjective genetic algorithm (MOGA) is used to vary library size and configuration simultaneously with other library properties. The result is a family of solutions that explores the tradeoffs in the objectives. This is achieved without the need to assign relative weights to the objectives. The user is then able to make an informed choice on an appropriate compromise solution. The method has been applied to two different virtual libraries: a two-component aminothiazole library and a four-component benzodiazepine library.
本文探讨了图书馆设计中的一个主要问题，即如何在与多样性、成本和类药物理化性质等其他属性同时，高效地优化图书馆规模（产品数量）和配置（每个位置的试剂数量）。这些目标往往相互冲突，例如，在尽量减少反应物数量的同时最大化多样性，因此给遗传算法和模拟退火等传统优化方法带来了困难。在此，使用多目标遗传算法（MOGA）来同时改变图书馆规模和配置以及其他图书馆属性。结果是一系列探索目标之间权衡的解决方案。这一过程无需为目标分配相对权重即可实现。然后，用户能够在合适的折衷解决方案上做出明智的选择。该方法已应用于两个不同的虚拟库：一个双组分氨基噻唑库和一个四组分苯二氮卓库。
