Premium
QSAR Models for Toxicity of Organic Substances to Daphnia magna Built up by Using the CORAL Freeware
Author(s) -
Toropova Alla P.,
Toropov Andrey A.,
Benfenati Emilio,
Gini Giuseppina
Publication year - 2012
Publication title -
chemical biology and drug design
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.59
H-Index - 77
eISSN - 1747-0285
pISSN - 1747-0277
DOI - 10.1111/j.1747-0285.2011.01279.x
Subject(s) - daphnia magna , quantitative structure–activity relationship , molecular descriptor , test set , set (abstract data type) , biological system , computer science , chemistry , mathematics , artificial intelligence , machine learning , biology , toxicity , organic chemistry , programming language
CORAL (CORrelations And Logic, http://www.insilico.eu/coral/) is a freeware available on the Internet. This freeware is designed to build up quantitative structure - property/activity relationships. The molecular structure for CORAL should be represented by the simplified molecular input line entry system (SMILES). Optimal descriptors calculated with SMILES are a mathematical function of the presence or absence of SMILES elements. The essence of this approach is the calculation of correlation weights for each element or combination of the elements by the Monte Carlo method. These coefficients serve to calculate the descriptors correlated with the endpoint for the training set, hoping that this correlation will also hold for the external test set. These descriptors can be improved by taking into account global physicochemical situations in molecules. An example of the physicochemical situation is the presence of oxygen and nitrogen. One can calculate these situations with SMILES and represent them by combining 0 (absence) and 1 (presence). The involving in the modelling of correlation weights of aforementioned physicochemical situations gave improvement in accuracy of models of toxicity to Daphnia magna for test set: n(test) = 75, r(2) = 0.7322, r(2) (pred) = 0.7193, r(2) (m) = 0.6549 (without correlation weights of the physicochemical situations); and n(test) = 75, r(2) = 0.7897, r(2) (pred) = 0.7790, r(2) (m) = 0.6850 (with aforementioned correlation weights of physicochemical situations).