Study On The Structure Activity Relationship Of Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitors

Epidermal growth factor receptor (EGFR) is one of the important targets for anti-cancer drug therapy, and it becomes a big focus of anticancer drug design at present. In this research, we mainly use the method of multiple linear regression (MLR), self-organizing map (SOM), support vector machine (SVM) and similarity retrieval to study the quantitative structure-activity relationship of epidermal growth factor receptor inhibitors. The topic of this study can be divided into three parts, as follows:The first part mainly research the qualitative classification of inhibitors and decoys of epidermal growth factor receptor using self-organizing map (SOM) and support vector machine (SVM). We collected 1248 inhibitors and 3093 decoys, then divided the data into a training set and a test set based on the random method. The structure parameters of ADRIANA.Code were calculated for each compound, and selected 13 structure parameters by using correlation analysis and stepwise. Finally, the classification models were built using the 13 selected structure parameters. The prediction accuracies for the models on the training and test sets are 98.48% and 96.33% for SOM,99.45% and 97.58% for SVM. In addition, we analyze the relationship between the structure of inhibitor and the bioactivity using molecular docking analysis.The second part is mainly on the quantitative prediction of the bioactivity of epidermal growth factor receptor inhibitor, and a series of prediction models were built through multiple linear regression (MLR) and support vector machine (SVM) method. The data was classified into two parts according to the determination method for the activity of the compounds:the data set F (793 compounds) based on fluorescence detection and the data set R (819 compounds) based on radioactive detection. Each data set was divided into a training set and a test set by random and self organizing map, respectively. Similarly, the structure parameters of ADRIANA.Code for each molecule were calculated, and the best structure parameters set was selected by using correlation analysis and stepwise regression. Then the prediction models were established based on the selected descriptors. For all the prediction models, the correlation coefficient square R2 on training set were higher than 0.70, the standard deviation (sd) were less than 0.71; the correlation coefficient square R2 on test sets were higher than 0.62, the standard deviation were less than 0.86.In the third part,2D and 3D structure similarity retrieval of compounds were carried out based on the epidermal growth factor receptor (EGFR) inhibitors.2D structure similarity retrieval was based on five fingerprints and three similarity coefficient methods.3D structure similarity retrieval was based on the 3D shape of molecules and Tanimoto coefficient. This study aimed to compare different method and 3D retrieval method to search aim compounds from database. As a result, two optimal two-dimensional search methods were obtained:Tanimoto-Path and Euclid-Path, their enrichment rate are 0.948, the enrichment of 3D retrieval rate was 0.879.