Vol. 5 Issue 3
Year:2016
Issue:Jun-Aug
Title:Linear Regression Analysis and Validation Studies of Insulin-Like Growth Factor (IGF-1) Receptor Inhibitors
Author Name:R. Rambabu and P. Srinivasa Rao
Synopsis:
To delineate the dependency of physico-chemical properties on activity of inhibitors, a python program was written to study the linear regression analysis on a set of Insulin Growth Factor – 1R inhibitors. About 87 IGF-1 receptor inhibitors were selected from published literature with few independent variables such as Molecular weight, Hydrogen bond donors, acceptors, logP and a number of freely rotatable bonds, 5- and 6- membered aromatic rings, Randic, Balaban indices, LUMO, HOMO, Dipole, Lipole and molar refractivity were selected. The relationship between dependent variable (log1/IC ) and independent variables was established by linear regression analysis using python code. A linear regression 50 2 analysis resulted in F-test: 8.812, r value: 0.794 and r value of 0.631, respectively. Inter-correlation between variables of the proposed regression model was checked to know about their independence. It is observed that compounds 13, 53 and 63 has standardized residuals -2.011, 2.309 and 2.227 respectively and can safely be excluded from the data set. Leverages and standardized residuals resulted in similar outcomes, but leverage analysis was able to find three more outlying data, finally 6 compounds were omitted from the dataset of 87 compounds, and the remaining 81 were divided as 76 molecule training set and a 5 molecule validation set. The predicted values of test set data (actual vs predicted) 2 2 when Equation (2) was applied and the regression coefficient (r ) obtained was 0.9686 and regression coefficient (r ) o passing through origin was plotted, which is 0.9326 within the limits. Regression plot between actual vs. predicted values 2 2 of compounds and vice versa of test data set showed r = 0.9686 and r = 0.9465, which suggests the predictive ability of o 2 the regression equation. From the analysis of test set, the regression equation is said to have predictive ability with R = cv,ext 2 2 2 2 2 2 2 0.99, R = 0.97, (R – R ) / R = 0.03 and (R – R ) / R = 0.02 and k = 1.01 and k' = 0.99 respectively.
Year:2016
Issue:Jun-Aug
Title:Linear Regression Analysis and Validation Studies of Insulin-Like Growth Factor (IGF-1) Receptor Inhibitors
Author Name:R. Rambabu and P. Srinivasa Rao
Synopsis:
To delineate the dependency of physico-chemical properties on activity of inhibitors, a python program was written to study the linear regression analysis on a set of Insulin Growth Factor – 1R inhibitors. About 87 IGF-1 receptor inhibitors were selected from published literature with few independent variables such as Molecular weight, Hydrogen bond donors, acceptors, logP and a number of freely rotatable bonds, 5- and 6- membered aromatic rings, Randic, Balaban indices, LUMO, HOMO, Dipole, Lipole and molar refractivity were selected. The relationship between dependent variable (log1/IC ) and independent variables was established by linear regression analysis using python code. A linear regression 50 2 analysis resulted in F-test: 8.812, r value: 0.794 and r value of 0.631, respectively. Inter-correlation between variables of the proposed regression model was checked to know about their independence. It is observed that compounds 13, 53 and 63 has standardized residuals -2.011, 2.309 and 2.227 respectively and can safely be excluded from the data set. Leverages and standardized residuals resulted in similar outcomes, but leverage analysis was able to find three more outlying data, finally 6 compounds were omitted from the dataset of 87 compounds, and the remaining 81 were divided as 76 molecule training set and a 5 molecule validation set. The predicted values of test set data (actual vs predicted) 2 2 when Equation (2) was applied and the regression coefficient (r ) obtained was 0.9686 and regression coefficient (r ) o passing through origin was plotted, which is 0.9326 within the limits. Regression plot between actual vs. predicted values 2 2 of compounds and vice versa of test data set showed r = 0.9686 and r = 0.9465, which suggests the predictive ability of o 2 the regression equation. From the analysis of test set, the regression equation is said to have predictive ability with R = cv,ext 2 2 2 2 2 2 2 0.99, R = 0.97, (R – R ) / R = 0.03 and (R – R ) / R = 0.02 and k = 1.01 and k' = 0.99 respectively.
No comments:
Post a Comment