QSAR studies on dipeptides based on a combinatorial MHDV-GA-MLR method

A com bi na torial method for es ti mat ing and pre dict ing the bi o log i cal ac tiv i ties of two sets of dipeptides, a set of 48 com pounds and an other set of 58, was de vel oped. The mo lec u lar ho lo graphic dis tance vec tor (MHDV) was em ployed to char ac ter ize the struc tures of the pep tide mol e cules. Pre lim i nary se lec tion of the MHDV descriptors was per formed based on the num ber of the mol e cules hav ing non-zero MHDV val u es. The fi nal op ti mal descriptors were com pleted by a ge netic al go rithm-based vari able se lec tion pro ce dure. Then the op ti mal descriptors are used to re late to the bi o log i cal ac tiv i ties of the pep tides us i ng the mul ti ple lin ear re gres sion (MLR) method. For two pan els of dipeptides, the cor re la tion co ef fi cient of es ti m a tions (r) are re spec tively 0.9651 for 48 pep tides and 0.936 for 58 pep tides, and the cor re la tion co ef fi cient of leave-one-out pre dic tions (q) are re spec tively 0.9452 and 0.9075. Mo lec u lar struc tural char ac ter iza tion (MSC) is an im por tant and in dis pens able tech nique in mod ern drug mo lec u lar de sign and the eval u a tion of phar ma co log i cal pro files. To study the bi o log i cal ac tiv i ties of the mol e cules and de velop the quan ti ta tive struc ture-activity re la tion ships (QSAR) so as to de sign fur ther novel mol e cules and eval u ate their bi o log i cal ac tiv i ties, it is es sen tial to de scribe the chem i cal struc tures of the mol e cules. The cur rent meth ods of MSC in clude two-dimensional top o log i cal descriptors, en er getic descrip tors, quan tum me chan i cal descriptors, and three-dimensional mo lec u lar field descriptors. 1