QSAR Studies On Some Drug Systems Using Novel Chemometric Algorithms

Quantitative structure-activity relationship (QSAR) is an important branch in chemometrics. Based on several novel algorithms proposed by our laboratory, some applications have been performed in the presented Master D. thesis. The first step in building a QSAR model is to calculate a great number of molecular descriptors, then the second is to select the descriptors most correlated to the studied object. But this reduction of variables often leads to the loss of information. Using Optimized Block-wise Variable Combination (OBVC) by Particle Swarm Optimization (PSO) based on PLS modeling to perform the optimized combination of variables and establish the regression model with the most related latent variables, the inhibitory activity of substituted bis[(acridine-4-carboxamide)propyl]methylamines and cyclooxygenase (COX) have been predicted. The result showed that modified PSO is a useful tool for searching optimized variable combination which converges quickly towards the optimal position. In QSAR studies, as the structural diversity in QSAR training set increases, there might be difficult to use a single linear model for the whole population of compounds of interest with a desired error level. To circumvent this problem, piecewise hyper-sphere modeling by particle swarm optimization (PHMPSO) was introduced to chemometrics. This method is to find multiple models by splitting the whole data set into subsets with desired linearity in each model. Using this algorithm, QSAR model of a series of 2-aryl(heteroaryl)-2,5-dihydropyrazolo-[4,3-c]quinolin-3-(3H)-ones (PQs) was constructed and the results were compared to those obtained by MLR in a single model and K-means clustering analysis in MLR modeling. It has been demonstrated that PHMPSO is a useful tool for improving the QSAR model. Through the analysis of the descriptors used in building models some information about the interaction between compounds and biomacromolecules can be achieved, which will guide the subsequent drug design and synthesis. Wavelength selection methods are variable selection techniques that allow calibration models to be constructed with a subset of spectral points instead of with the whole spectrum. In the appended wok, we collected the Near-Infrared absorbance spectra of the plasma samples spiked with different amount of caffeine. All the spectral data were treated by a version of modified PSO to select the optimal spectral intervals. Calibration models were constructed separately on the selected spectral intervals and the full spectrum using PLS. Comparing the two models shows that the spectra intervals selected by modified PSO can be helpful in PLS modeling and it also demonstrated that the modified PSO is a useful tool for wavelength selection.