Associative Neural Network
An associative neural network (ASNN) is an ensemble-based method inspired by the function and structure of neural network correlations in brain. The method operates by simulating the short- and long-term memory of neural networks. The long-term memory is represented by ensemble of neural network weights, while the short-term memory is stored as a pool of internal neural network representations of the input pattern. The organization allows the ASNN to incorporate new data cases in short-term memory and provides high generalization ability without the need to retrain the neural network weights. The method can be used to estimate a bias and the applicability domain of models. The applications of the ASNN in QSAR and drug design are exemplified.
KeywordsEnsemble networks memory drug design LIBRARY mode
Chemical company, www.basf.com.
Cascade correlation neural network
Central processing unit
k nearest neighbors
- LIBRARY mode
An operational mode of the ASNN when new compounds are used to correct neural network ensemble predictions without changing neural network weights (see Eq. 6)
- log D
The same as log P but for ionized compounds (usually measured at a specific pH)
- log P
1 Octanol/water partition coefficient
- log S
Aqueous solubility of compounds
Nuclear magnetic resonance
Physical properties database 
- “nova” set
- “star” set
Quantitative structure-activity relationship studies
Root mean squared error
University of California, Irvine
This study was supported by the Virtual Computational Chemistry Laboratory grant INTAS INFO-00363. I thank Philip Wong for his useful comments and remarks.
- 1.Fuster JM (1995) Memory in the cerebral cortex. MIT Press, Cambridge, MA,.Google Scholar
- 2.Tetko V (2001) Associative Neural Network, CogPrints Archive, cog00001441.Google Scholar
- 6.Tetko IV, Livingstone DJ, Luik AI (1995) Neural network studies, 1. Comparison of overfitting and overtraining. J Chem Inf Comput Sci 35:826–833.Google Scholar
- 9.Tetko IV, Villa AEP (1995) In Unsupervised and supervised learning: cooperation toward a common goal, In: ICANN'95, international conference on artificial neural networks NEURONIMES'95, Paris. EC2 & Cie, Paris, France, pp 105–110.Google Scholar
- 14.Fahlman S, Lebiere C (1990) The cascade-correlation learning architecture. NIPS 2:524–532.Google Scholar
- 15.Blake EK, Merz C (1998) UCI repository of machine learning databases, available www.ics.uci.edu/∼mlearn/MLRepository.html.
- 18.Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1994) Numerical recipes in C (2nd edn). Cambridge University Press, New York, p. 998.Google Scholar
- 19.The Physical Properties Database (PHYSPROP), Syracuse Research Corporation, available www.syrres.com, accessed December 20, 2006.
- 25.Kandemirli F, Shvets N, Unsalan S, Kucukguzel I, Rollas S, Kovalishyn V, Dimoglo A (2006) The structure-antituberculosis activity relationships study in a series of 5-(4-aminophenyl)-4-substituted-2,4-dihydro-3h-1,2,4-triazole-3-thione derivatives. A combined electronic-topological and neural networks approach. Med Chem 2:415–422.CrossRefPubMedGoogle Scholar
- 27.Ajmani S, Tetko IV, Livingstone DJ, Salt D (2005) A comparative study of neural network architectures for QSAR., In: Aki(Sener) E, Yalcin I (eds) QSAR and molecular modelling in rational design of bioactive molecules, Computer Aided Drug Design and Development Society in Turkey, Istanbul, pp. 183–184.Google Scholar
- 29.Tetko IV, Solov'ev VP, Antonov AV, Yao X, Doucet JP, Fan B, Hoonakker F, Fourches D, Jost P, Lachiche N, Varnek A (2006) Benchmarking of linear and nonlinear approaches for quantitative structure-property relationship studies of metal complexation with ionophores. J Chem Inf Model 46:808–819.CrossRefPubMedGoogle Scholar
- 30.Vapnik VN (1998) Statistical leaning theory. Wiley, New York,.Google Scholar
- 33.Tetko IV, Livingstone DJ (2007) Rule-based systems to predict lipophilicity. In: Testa B, van de Waterbeemd H (eds) Comprehensive medicinal chemistry II: in silico tools in ADMET, vol. 5. Elsevier, Oxford, UK, pp 649–668.Google Scholar
- 34.Poda GI, Tetko IV, Rohrer DC (2005) Towards predictive ADME profiling of drug candidates: lipophilicity and solubility. In: 229th American Chemical Society national meeting and exposition, ACS, San Diego, CA, p. MEDI 514.Google Scholar
- 38.Tetko IV, Tanchuk VY (2005) ALOGPS (www.vcclab.org) is a free on-line program to predict lipophilicity and aqueous solubility of chemical compounds. In: 229th American Chemical Society national meeting and exposition, ACS, San Diego, CA pp. U608–U608.
- 39.Tetko IV (2005) Encoding molecular structures as ranks of models: a new secure way for sharing chemical data and development of ADME/T models. In 229th American Chemical Society national meeting and exposition, San Diego, CA, pp. U602–U602.Google Scholar
- 40.Tetko IV, Bruneau P, Mewes HW, Rohrer DC, Poda GI (2006) Can we estimate the accuracy of ADMET predictions? In: 232th American Chemical Society national meeting, San Francisco.Google Scholar
- 42.Tetko IV (2006) In estimation of applicability domain of a model for toxicity against T. pyriformis using ALOGPS logP. Workshop on ranking methods, Verbania, Italy, October 2–3.Google Scholar
- 44.Kier LB, Hall LH (1999) Molecular structure description: the electrotopological state. Academic Press, London, p. 245.Google Scholar
- 45.Hall LH, Kier LB (1995) Electrotopological state indices for atom types—a novel combination of electronic, topological, and valence state information. J Chem Inf Comput Sci 35:1039–1045.Google Scholar
- 50.Härdle W (1990) Smoothing techniques with implementation in S. Springer-Verlag, New York.Google Scholar