Title:QSAR Analysis of Some Antagonists for p38 map kinase Using Combination of Principal Component Analysis and Artificial Intelligence
Volume: 18
Issue: 8
Author(s): Elham Doosti and Mohsen Shahlaei
Affiliation:
Keywords:
ANN, p38 MAP kinase, PCA, QSAR.
Abstract: Quantitative relationships between structures of a set of p38 map kinase inhibitors and their
activities were investigated by principal component regression (PCR) and principal componentartificial
neural network (PC-ANN). Latent variables (called components) generated by principal
component analysis procedure were applied as the input of developed Quantitative structure- activity
relationships (QSAR) models. An exact study of predictability of PCR and PC-ANN showed that the
later model has much higher ability to calculate the biological activity of the investigated molecules.
Also, experimental and estimated biological activities of compounds used in model development step have indicated a
good correlation.
Obtained results show that a non-linear model explaining the relationship between the pIC50s and the calculated principal
components (that extract from structural descriptors of the studied molecules) is superior than linear model.
Some typical figures of merit for QSAR studies explaining the accuracy and predictability of the suggested models were
calculated.
Therefore, to design novel inhibitors of p38 map kinase with high potency and low undesired effects the developed QSAR
models were used to estimate biological pIC50 of the studied compounds.