8. QSPR and ANN Studies on Prediction of Aqueous Solubility of Heterogeneous Set of Organic Compounds

Bruno Louis^a, Jyoti Singh^b, Basheerulla Shaik^­b,
Vijay K. Agrawal^­b*, Padmakar V. Khadikar^c
aDepartment of Pharmacy,PO Box 38, Sultan Qaboos University Hospital, Al Khod Muscat 123,Oman,
E-mail: louisb4425@yahoo.com
^­b*QSAR and Computer Chemical Laboratories, A.P.S. University, Rewa-486 003, India E-mail: apsvkal@yahoo.co.in
^c Research Division, Laxmi Fumigation and Pest Control Pvt. Ltd.
3 Khatipura, Indore-452 007, India  E-mail: pvkhadikar@rediffmail.com

Abstract: The aqueous solubility of 399 heterogeneous organic compounds was predicted by quantitative structure–property relationship (QSPR) method. In this work, only topological descriptors and indicator parameters were used.  The topological descriptors used are whole molecular structural descriptors derived from theoretical molecular calculations (not atom or bond count). The multiple linear regression (MLR) (for 398 compounds R = 0.898) and artificial neural network (ANN) techniques were used to build linear and nonlinear models, respectively.  In this work the proposed QSPR models, both by MLR and ANN, contain identical descriptors. Comparison of these two methods reveals that those obtained by the ANN model are better.

 Keywords: Artificial neural network, solubility, topological descriptors, Regression analysis, QSPR

<<<