Toward generating simpler QSAR models: Nonlinear multivariate regression versus several neural network ensembles and some related methods

Journal of Chemical Information and Computer Sciences

Volumn 43, Issue 4, 2003, Pages 1094-1102

  Lučić, Bono ^a   Nadramija, Damir ^b   Basic, Ivan ^b   Trinajstic, Nenad ^a  

^a RUDER BO KOVI INSTITUTE   (Croatia)

^b Pharmaceutical Industry   (Croatia)

Author keywords

[No Author keywords available]

Indexed keywords

BACKPROPAGATION; DERIVATIVES; GENETIC ALGORITHMS; LEAST SQUARES APPROXIMATIONS; MATHEMATICAL MODELS; REGRESSION ANALYSIS;

DATA SETS;

NEURAL NETWORKS;

EID: 18344410789     PISSN: 00952338     EISSN: None     Source Type: Journal    
DOI: 10.1021/ci025636j     Document Type: Article

References (22)

1. Lučić, B.; Trinajstić, N. Multivariate Regression Outperforms Several Robust Architectures of Neural Networks in QSAR Modeling. J. Chem. Inf. Comput. Sci. 1999, 39, 121-132.
2. Lučić, B.; Trinajstić, N.; Sild, S.; Karelson, M.; Katritzky, A. R. A New Efficient Approach for Variable Selection Based on Multiregression: Prediction of Gas Chromatographic Retention Times and Response Factors. J. Chem. Inf. Comput. Sci. 1999, 39, 610-621.
3. Lučić, B.; Amić, D.; Trinajstić, N. Nonlinear Multivariate Regression Outperforms Several Concisely Designed Neural Networks in QSAR Modeling. J. Chem. Inf. Comput. Sci. 2000, 40, 403-413.
4. Basak, S. C.; Gute, B. D.; Lučić, B.; Nikolić, S.; Trinajstić, N. A Comparative QSAR Study of Benzamidines Complement-Inhibitory Activity and Benzene Derivatives Acute Toxicities. Comput. Chem. 2000, 24, 181-191.
5. Tetko, I. V.; Alessandro Villa, A. E. P.; Livingston, D. J. Neural Network Studies. 1. Comparison of Overfilling and Overtraining. J. Chem. Inf. Comput. Sci. 1995, 35, 826-833.
6. Tetko, I. V.; Alessandro Villa, A. E. P.; Livingston, D. J. Neural Network Studies. 2. Variable Selection. J. Chem. Inf. Comput. Sci. 1996, 36, 794-803.
7. Kovalishyn, V. V.; Tetko, I. V.; Alessandro Villa, A. E. P.; Livingston, D. J. Neural Network Studies. 3. Variable Selection in the Cascade-Correlation Learning Architecture. J. Chem. Inf. Comput. Sci. 1998, 38, 651-659.
8. Hoffmann, R.; Minkin, V. I.; Carpenter, B. K. Ockham's Razor and Chemistry. Bull. Soc. Chim. Fr. 1996, 133, 117-130.
9. Cerius2; Accelrys: 9685 Scranton Road, San Diego, CA, 92191.
10. Rogers, D.; Hopfinger, A. J. Application of Genetic Function Approximation to Quantitative Structure-Activity Relationships and Quantitative Structure-Property Relationships. J. Chem. Inf. Comput. Sci. 1994, 34, 854-866.
11. Nakao, H.; Arakawa, M.; Nakamura, T.; Fukushima, M. Antileukemic Agents. II. New 2,5-bis(l-aziridinyl)-p-benzoquinone Derivatives. Chem. Pharm. Bull. 1972, 20, 1968-1979.
12. Liu, Q.; Hirono, S.; Moriguchi, I. Comparison of Functional-Link Net and Generalised Delta Rule Net in Quantitative Structure-Activity Relationship Studies. Chem. Pharm. Bull. 1992, 40, 2962-2969.
13. Aoyama, T.; Suzuki, Y.; Ichikawa, H. Neural Networks Applied to Quantitative Structure-Activity Relationship Analysis. J. Med. Chem. 1990, 33, 2583-2590.
14. Hirst, J. D.; King, R. D.; Sternberg, M. J. E. Quantitative Structure-Activity Relationships by Neural Networks and Inductive Logic Programming. 1. The Inhibition of Dihydrofolate Reductase by Pyrimidines. J. Comput.-Aided Mol. Des. 1994, 8, 405-420.
15. Selwood, D. L.; Livingstone, D. J.; Comley, J. C. W.; O'Dowd, A. B.; Hudson, A. T.; Jackson, P.; Jandu, K. S.; Rose, V. S.; Stables, J. N. Structure-Activity Relationships of Antifilarial Antimycin Analogues: A Multivariate Pattern Recognition Study, J. Med. Chem. 1990, 33, 136-142.
16. Dunn, W.; Rogers, D. Genetic Partial Least Squares in QSAR. In Genetic Algorithms in Molecular Modeling; Devillers, J., Ed.; Academic Press: London, 1996; pp 109-130.
17. Kubinyi, H. Evolutionary Variable Selection in Regression and PLS Analyses. J. Chemometrics 1996, 10, 119-133.
18. So, S.-S.; Karplus, M. Evolutionary Optimization in Quantitative Structure-Activity Relationship: An Application of Genetic Neural Network. J. Med. Chem. 1996, 39, 1521-1530.
19. Whitley, D. C.; Ford, M. G.; Living stone, D. J. Unsupervised Forward Selection: A Method for Eliminating Redundant Variables. J. Chem. Inf. Comput. Sci. 2000, 40, 1160-1168.
20. Waller, C. L.; Bradley, M. P. Development and Validation of a Novel Variable Selection Techniques with Application to Multidimensional Quantitative Structure-Activity Relationships Studies. J. Chem. Inf. Comput. Sci. 2000, 39, 345-355.
21. For example, in ref 6 the authors pointed on page 795 (section "ANN implementation") that single ANN (Artificial Neural Network) with 10 neurons in one hidden layer were used in the computation. In addition, bias neurons were used both on the input and on the hidden layer. It means that single ANN contained up to 77 weights for only five inputs. Because at least 100 ANNs were calculated and used in each ensemble a huge number of more than 7000 weights (i.e. optimized parameters) were used in each NNE model. It is important to point out that, for some data sets, even 500 ANNs were used in the ensemble.
22. Lučić, B.; Bašic, I.; Nadramija, D.; Miličević, A.; Trinajstić, N.; Suzuki, T.; Petrukhin, R.; Karelson, M.; Katritzky, A. R. Correlation of liquid viscosity with molecular structure for organic compounds using different variable selection methods. Arkivoc 2002, (IV), 45-59 (http;// www.arkat-usa.org/ark/journal/2002/Sunko/DS-381D/DS-38 1D.pdf).