Exploring QSARs for inhibiting activity of epidermal growth factor receptor (EGFR) tyrosine kinase by MLR and PC-ANN
Date
2017-05-06
Authors
Manal Mohammad Wael Abdul Hafez Mohtaseb
منال محمد وائل عبد الحافظ المحتسب
Journal Title
Journal ISSN
Volume Title
Publisher
Al-Quds University
Abstract
Quantitative structure- activity relationship (QSAR) study was preformed to understand the activity of a set of 113 compounds of Epidermal Growth Factor Receptor (EGFR) inhibitors.
QSAR models were developed using multiple linear regression (MLR) as linear method. While principle component- artificial neural network (PC-ANN) modeling method was performed as nonlinear method.
The MLR resulted with models (12-23) which have coefficient of determination (R2)>0.6, the best model (model 23) resulted with correlation coefficient (R) = 0.878, coefficient of determination (R2) =0.771, and adjusted coefficient of determination (R2adj) =0.719. Cross validation leave one out (LOO) and leave many out (LMO) were performed on the resulted MLR models, models 19-23 showed a good predictive power. After that principle component analysis (PCA) performed to divide the data into three data sets. Then the ANN performed on the chosen models (19-23) from leave one out (LOO) and leave many out (LMO) cross validation. ANN resulted models were validated through randomization test.
The best ANN model with good predictive power was model 19 with R=0.812 for the test set.