PLEASE CLICK ON PDF TO VIEW THE ABSTRACT.

IN RECENT YEARS, HETEROGENIZATION OF ORGAN OCATALYTIC SYSTEMS HAS BEEN ACHIEVED BY IMMOBILIZATION OF HOMOGENEOUS CATALYSTS ON VARIOUS TYPES OF ORGANIC AND INORGANIC SUPPORTS. ONE OF THIS SURFACE THAT ACT AS A SUPPORT IS CELLULOSE. CELLULOSE, BEING THE MOST ABUNDANT NATURALLY OCCURRING POLYMER ON EARTH, IS A HIGHLY INTERESTING MATERIAL DUE TO ITS RENEWABILITY, LOW PRICE, HIGH AVAILABILITY, AND GOOD MECHANICAL PROPERTIES. THE INTEREST IN CELLULOSE HAS REAWAKENED DUE TO THE CONCERN FOR THE ENVIRONMENTAL IMPACT OF SYNTHETIC POLYMERS AND FIBRES, AND AS THE PRICE OF THE RAW MATERIAL FOR PRODUCTION OF SYNTHETIC POLYMERS IS RAPIDLY INCREASING. SINCE CELLULOSE IS BOTH BIORENEWABLE AND BIODEGRADABLE, SEVERAL STUDIES HAVE FOCUSED ON THE MODIFICATION OF CELLULOSE, TO INVESTIGATE IT AS A REPLACEMENT MATERIAL FOR SYNTHETIC POLYMERS [1,2].EPOXIDES ARE IMPORTANT CLASS OF INDUSTRIAL CHEMICALS THAT HAVE BEEN USED AS CHEMICAL INTERMEDIATES AND CAN BE OPENED UNDER A VARIETY OF CONDITIONS [4].IN THIS WORK, THE CATALYTIC ACTIVITY OF MN (SALEN) AND MN (SALEN)-CELLULOSE IN THE NUCLEOPHILIC RING-OPENING OF EPOXIDES AND IN THE ETOH AS SOLVENT WITH KSCN, NASCN AND THIOUREA AS NUCLEOPHIL IS REPORTED.

QSAR STUDY FOR MODELING AND PREDICTING ANTI-HYPERTENSIVE ACTIVITY OF A SERIES OF 65 NON-UREA INHIBITORS OF SOLUBLE EPOXIDE HYDROLASE APPLIED BY MEANS OF MULTIPLE LINEAR REGRESSION (MLR), PARTIAL LEAST SQUARE (PLS), PRINCIPAL COMPONENTS REGRESSION (PCR), AND ARTIFICIAL NEURAL NETWORK (ANN) TECHNIQUES. FIRST, THIS DATA SET WAS DIVIDING INTO TWO GROUPS INCLUDING TRAINING (45 MOLECULES) AND TEST (20 MOLECULES) SETS. THEN A LARGE NUMBER OF DESCRIPTORS WERE CALCULATED USING HYPERCHEM AND DRAGON PROGRAMS FOR THESE COMPOUNDS. A SUITABLE NUMBER OF THESE DESCRIPTORS SELECTED USING STEPWISE MULTIPLE LINEAR REGRESSION (MLR) TECHNIQUE. THE RESULTS OBTAINED FROM THIS METHOD INDICATES THAT ANTI-HYPERTENSIVE ACTIVITY OF STUDIED COMPOUNDS DEPENDS ON DIFFERENT PARAMETERS CONTAINING TOPOLOGICAL, AUTOCORRELATION, 3D MORSE AND ATOM CENTERED FRAGMENTS DESCRIPTORS. THEREFORE, THESE DESCRIPTORS HAVE USED AS INPUT FOR THE ANN MODELING METHOD. THE BEST ARTIFICIAL NEURAL NETWORK IS A PROGRESSIVE NETWORK WITH A BACKWARD DISTRIBUTION OF ERRORS THAT INSTRUCTED BY LEVEN BERG-MARAQUARDT ALGORITHM (LM). FINALLY, OBTAINED RESULTS USING ALL MODELING METHODS COMPARED. CORRELATION COEFFICIENT BETWEEN EXPERIMENTAL AND CALCULATED ANTI-HYPERTENSIVE ACTIVITY OF STUDIED COMPOUNDS AND STANDARD ERROR FOR THE TRAINING AND TEST SETS USING MLR, PLS AND PCR METHODS ARE SHOWN IN TABLE 1. IN THE CASE OF ANN, DATA SET DIVIDED INTO THREE GROUPS OF TRAINING, TEST, AND VALIDATION. THE BEST RESULTS OBTAINED USING ANN MODEL THAT SHOWN IN TABLE 1.