摘要
计算机辅助分子设计是研发浮选药剂的重要工具,计算化学结合机器学习关联理论与现实,可提高药剂设计的精准度。实现数据驱动的分子设计很大程度依赖科学的定量结构-活性关系(QSAR),将捕收剂分子结构输入计算程序,输出各种化学性质描述符,同捕收剂试验特性的响应建立映射关系,正逐渐成为现今捕收剂高通量分子设计与筛选的一套范式。QSAR的关键在于描述符的筛选与映射的拟合,围绕这两个问题,本文系统性梳理了构建浮选捕收剂QSAR的流程,阐述了获取并处理捕收剂结构-性质-响应数据的方法;根据计算化学理论描述捕收剂亲固性特征与疏水性特征,分析亲固性与疏水性机理;介绍了通过机器学习算法筛选高相关性的描述符并训练预测模型的方法。以此为基础,展望数据驱动的浮选捕收剂分子设计的未来发展方向。
Computer-aided molecular design is an important tool in the development of flotation reagent,and the combination of computational chemistry with machine learning to correlate theory and reality can improve the accuracy of reagent design.The realization of data-driven molecular design relies heavily on the scientific quantitative structure-activity relationship(QSAR),which is a paradigm for high-throughput molecular design and screening of collectors by inputting the molecular structure of collectors into a computational program,outputting various chemical property descriptors,and mapping them with the responses of collectors'experimental properties.The key of QSAR lies in the screening of the descriptors and fitting of the mapping.Focusing on these two issues,this paper systematically combs through the process of constructing a QSAR for flotation collectors,describes the methods of obtaining and processing structure-property-response data of collectors,describes the characteristics of collectors in terms of their affinity and hydrophobicity according to the theory of computational chemistry,and analyzes the mechanism of affinity and hydrophobicity,and introduces the screening of high relevance descriptors and training of descriptors by machine learning algorithm.Based on this,we look forward to the future development of data-driven molecular design of flotation collectors.
作者
叶子涵
刘胜
陈伟
刘广义
YE Zi han;LIU Sheng;CHEN Wei;LIU Guangyi(State Key Laboratory of Vanadium and Tit anium Resources Comprehensive Utilization,College of Chemistry and Chemical Engineering,Central South University,Changsha 410083,China)
出处
《有色金属(选矿部分)》
2025年第2期8-32,共25页
Nonferrous Metals(Mineral Processing Section)
基金
国家重点研发计划项目(2022YFC2904601)。
关键词
浮选捕收剂
QSAR
计算化学
描述符
机器学习
flotation collectors
QSAR
computational chemistry
descriptors
machine learning
