An improved QSPR study of the toxicity of aliphatic carboxylic acids using genetic algorithm

作者：Kompany Zareh M^*

来源：Medicinal Chemistry Research, 2009, 18(2): 143-157.

DOI：10.1007/s00044-008-9114-3

• QSTR
• Genetic algorithm
• Aliphatic carboxylic acids
• Toxicity

Toxicity of 38 aliphatic carboxylic acids was studied by using a quantitative structure-activity relationship (QSAR) method and the following molecular and topological properties: acidic dissociation constant, electrotopological states, molar refractivity, refraction index, log K(ow), E(LUMO), surface tension, and polarizability. Application of realistic molecular quantities that can express the toxicities of the molecules is the idea behind the choice of these topological and physicochemical descriptors. To obtain a proper model for relating the toxicity values of aliphatic carboxylic acids to their structural properties a genetic algorithm (GA) coupled with the partial least-squares (PLS) method were utilized. All different possible relations between toxicity [log (IGC(50)(-1))] and the molecular properties up to forth order were examined, and a group of multiple regression models with high fitness scores were generated. Acidic dissociation constant, surface tension, log Kow, and the electrotopological state of the methyl group were selected as significant descriptors, using a backward elimination method on the top descriptors obtained from the genetic algorithm. The selected descriptors and the orders chosen for them are shown to be good for assessing the toxicity of aliphatic carboxylic acids. It was also verified that, in equations in several variable, using higher-order relationships (i.e., square, cubic, and/or quadratic) gives an excellent accuracy.

• 出版日期2009-3