摘要
选择100个化合物作为数据集,随机选取其中80个为训练集,其他分子为验证集,并为每个化合物分子计算了30个参数.通过采用五种不同多元线性回归分析方法对其训练模拟,建立了数学模型,并用验证集检验了所建模型的预测能力.结果发现向后筛选法为最优小肠吸收建模方法.由该法所建模型的统计结果良好(R2>0.80),应用于验证集时也表现出较强预测能力.该模型确定了对小肠吸收影响较大的分子参数,有助于指导进一步的新药筛选和开发.
To quantitatively predict the fraction absorption of drugs of human intestine and determine the optimal regression method, a dataset composed of 100 diversified compounds,where 80 compounds served as training set and the rest ones as test set, was studied by several multivariate linear regression analysis methods. For each molecule, 30 molecular indices were calculated, resulting in a model with satisfactory statistical results （R^2〈0. 80） and proper predictability validated by the test set. From the analysis of the model, those key descriptors largely influencing the intestinal absorption of the molecules were identified,and Backword regression analysis was found to be the optimal regression method compared with the others. All these are valuable and helpful for aiding further screening and development of orally administered drugs.
出处
《分子科学学报》
CAS
CSCD
2007年第4期286-291,共6页
Journal of Molecular Science
基金
大连理工大学-中科院大连化学物理研究所合作科研探索基金资助项目
关键词
小肠吸收
分子参数
多元线性回归
intestinal absorption
molecular indices
multivariate linear regression analysis
