摘要
充分利用药物大数据平台和临床资源,运用数据分析方法预测抗乳腺癌候选药物的ADMET性质和抗乳腺癌活性,为实验室研制抗乳腺癌新药过程提供参考方向。针对1974种化合物的分子描述符变量数据,分别构建以ADMET性质和pIC 50值为因变量的随机森林预测模型,模型的预测精度分别为88.7%和91.3%。基于随机森林模型求得的重要影响因子贡献率确定出4个变化显著的共同重要影响因子的取值范围,分别为MLFER_BH(0.56,2.65)、MLFER_S(1.30,4.41)、WTPT-5(0.00,10.01)和SdssC(-1.92,2.76),对实现抗乳腺癌药物的优化具有指导意义。
By making full use of big pharmaceutical data platforms and clinical resources,we used data analysis methods to predict ADMET properties and anti-breast cancer activity of anti-breast cancer drug candidates.It provided a reference for the process of developing new anti-breast cancer drugs in the laboratory.Random forests were constructed for the molecular descriptor variable data of 1974 compounds,and the dependent variables of prediction models were ADMET properties and pIC50 values.The prediction accuracies of the models were 88.7%and 91.3%respectively.Based on the random forest model,we obtained the contribution rate of important impact factors.Then we established the ranges of four common significant influencers that varied significantly.They were MLFER_BH(0.56,2.65),MLFER_S(-1.30,4.41),WTPT-5(-0.00,10.01)and Sdssc(-1.92,2.76).The result was instructive to optimize the anti-breast cancer drugs.
作者
汤仕星
曾莹
TANG Shixing;ZENG Ying(School of Science,Hubei Univ.of Tech.,Wuhan 430068,China)
出处
《湖北工业大学学报》
2023年第1期111-115,120,共6页
Journal of Hubei University of Technology
基金
湖北省教育厅人文社科项目(19Q053)
湖北工业大学横向项目(2017124)。
