|
L. D. Grigoreva, V. Y. Grigorev, A. V. Yarkov
Outlier
detection for QSAR modeling of biological activity of chemicals on the basis of
"structure – activity – similarity" maps
Abstract
A new method of
outlier detection for QSAR was developed on the basis of “structure – activity
– similarity” (SAS) maps analysis. It consists in an empirical
assessment of the probability of appearance of a chemical compound in any area
of the SAS. It is proposed to consider compounds with the maximum probability
of appearance in the activity cliffs and a minimal probability in the smooth
region as outlier. The proposed method can also be used in the field of
medicinal chemistry to search for new biologically active compounds.
Key words: QSAR,
SAR, outlier, Structure – Activity – Similarity
maps.
Copyright (C) Chemistry Dept., Moscow State University, 2002
|
|
