【作者】 陈晓婕； 郑明月； 罗小民； 朱维良； 蒋华良； 陈凯先；

【机构】 中国科学院上海药物研究所新药研究国家重点实验室,药物发现与设计中心；

【摘要】 <正>Here we present several computational models that reveal the prediction abilities in absorption by the human intestine.We used a human intestinal absorption database contains 552 compounds to construct quantitative models and classification models for absorption prediction.The descriptors calculation and model construction were performed in Discovery Studio and Pipeline Pilot software packages.Genetic Function Approximation（GFA） and Recursive Partitioning（RP） served as two computational methods in our work.The best GFA model for HIA achieves an R² of 0.70 and an RMSE（root mean square error） of 15.96%for the training set.The predictive power of the model was further evaluated by using a test set of 98 compounds.The GFA model predicted the HIA with R²_test =0.70 and RMSE=12.63%for the test set.A qualitative HIA model was presented in RP tree form.In a further receiver operating characteristic（ROC） analysis,it gave a significant area under curve（AUC） value over 0.98,indicating a satisfactory predicting power.更多还原

【Abstract】 Here we present several computational models that reveal the prediction abilities in absorption by the human intestine.We used a human intestinal absorption database contains 552 compounds to construct quantitative models and classification models for absorption prediction.The descriptors calculation and model construction were performed in Discovery Studio and Pipeline Pilot software packages.Genetic Function Approximation（GFA） and Recursive Partitioning（RP） served as two computational methods in our work.The best GFA model for HIA achieves an R² of 0.70 and an RMSE（root mean square error） of 15.96%for the training set.The predictive power of the model was further evaluated by using a test set of 98 compounds.The GFA model predicted the HIA with R²_test =0.70 and RMSE=12.63%for the test set.A qualitative HIA model was presented in RP tree form.In a further receiver operating characteristic（ROC） analysis,it gave a significant area under curve（AUC） value over 0.98,indicating a satisfactory predicting power.更多还原