The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	QSAR modeling of chemical penetration enhancers using novel replacement algorithms
Author:	Qiu, Hui
View Online:	njit-etd2013-042 (xii, 55 pages ~ 0.8 MB pdf)
Department:	Department of Chemical, Biological and Pharmaceutical Engineering
Degree:	Master of Science
Program:	Pharmaceutical Bioprocessing
Document Type:	Thesis
Advisory Committee:	Simon, Laurent (Committee chair) Armenante, Piero M. (Committee member) Venanzi, Carol A. (Committee member) Loney, Norman W. (Committee member)
Date:	2013-05
Keywords:	Transdermal delivery Chemical penetration enhancers Chemical penetration enhancer structures
Availability:	Unrestricted
Abstract:	The applications of transdermal delivery are limited because of the resistance of the skin to drug diffusion. Only potent drugs, with molecular weight less than 500 Da, are suitable to cross the skin barrier. Chemical Penetration Enhancers (CPEs) are used to promote the absorption of solutes across the dermal layers. In this investigation, a Quantitative Structure-Activity Relationship (QSAR) model is applied to relate chemical penetration enhancer structures with the flux enhancement ratio through a statistical approach. A database, consisting of 61 non-polar CPEs, is selected for the study. Each compound is represented by 777 QSAR descriptors, which encode the physical characteristics of the CPE and its structure. Selection replacement techniques are used to choose the eight most important descriptors. The enhancement ratio, an evaluation of the effect of the CPE, correlates well with this subset of features. The QSAR model can be adopted to predict factors that need to be adjusted to improve permeation of the drug through the skin. Three QSAR models are developed using different algorithms: forward stepwise regression (FSR), replacement (RM) and enhanced replacement (ERM) techniques. The first two methods yield equations with poor predictive power. The enhanced replacement method gives the best results, which meet cross-validation criteria: q2 = 0.79, 0.63 and 0.76 for the training set, test set and combined data, respectively. These results meet the predetermined criteria.