QSAR modeling on the basis of 3D descriptors representing the electrostatic molecular surface (ambergris fragrances)

Solving the Structure-Property Problem Using k-NN Classification

“…Comparison of the quality of the k-NN classification using the algorithm of selection of descriptors (method 1) , based on all features (method 2) and based on factors (method 3) was conducted on a set of amber odorants (low molecular weight compounds with amber scent), consisting of 129 molecules [2]. 8 descriptors matrixes were used.…”

“…Solution of this problem can be divided into two stages: choosing the description of molecular graphs and the construction of the classifying function. structural descriptors (features) [1] -pairs and triples of singular points [2,3] defined on a triangulated molecular surface of the compound were selected for description and singular points, defined on molecular graphs.The structural character spectrum of the molecular graph represents the number of the molecular fragments repetitions in a molecular graph by complete enumeration of all pairs, triples, quadruples of the singular points [1,4,5]. As a result, the number of descriptors obtained is very large (about 1000 -10000) and we must choose how to use them.…”