Computational neural network predictive modeling

Goll, E.S. and Jurs, PC. (1999a). Prediction of the Normal Boiling Points of Organic Compounds from Molecular Structures with a Computational Neural Network Model. J.Chem.Inf.Com-put.Sci., 39,914-983. 

Models can be generated using stepwise addition multiple linear regression as the descriptor selection criterion. Leaps-and-bounds regression [10] and simulated annealing (ANNUN) can be used to find a subset of descriptors that yield a statistically sound model. The best descriptor subset found with multiple linear regression can also be used to build a computational neural network model. The root mean square (rms) errors and the predictive power of the neural network model are usually improved due to the higher number of adjustable parameters and nonlinear behavior of the computational neural network model. 

The multiple linear regression models are validated using standard statistical techniques. These techniques include inspection of residual plots, standard deviation, and multiple correlation coefficient. Both regression and computational neural network models are validated using external prediction. The prediction set is not used for descriptor selection, descriptor reduction, or model development, and it therefore represents a true unknown data set. In order to ascertain the predictive power of a model the rms error is computed for the prediction set. 

Compound used in the cross-validation set for computational neural networks. Compound used in the prediction set for regression and neural network models. 

The five descriptors found in this model were then fed to a computational neural network in an attempt to improve the predictive ability. The program ANN was used to optimize the starting weights and biases. The quality of the model was assessed by calculating the residuals [actual-predicted values of -logCLCso)] of the prediction set compounds. 

Goll, E.S. and Jurs, P.C. (1999a) Prediction of the normal boiling points of organic compounds from molecular structures with a computational neural network model./. Chem. Inf. Comput. Sci., 39,974— 983. 

Mattioni, B.E. and Jurs, P.C. (2003) Prediction of dihydrofolate reductase inhibition and selectivity using computational neural networks and linear discriminant analysis. /. Mol. Graph. Model., 21, 391-419. 

The preceding discussions show that statistical and computer neural network modeling can be applied to the GAP to obtain models that can be used to predict performance of the process in the forward and reverse modes. The neural network model predicts the performance to within an average of 12% for the forward and 35% for the reverse modes, making it a useful model for expanding the GAP technology to other systems without the need to perform large and expensive experiments. 

Johnson, S. R. and Jurs, P. C. (1997) Prediction of acute mammalian toxicity from molecular structure for a diverse set of substituted anilines using regression analysis and computational neural networks. Comput. Assist. Lead Find. Optim., [Eur. Symp. Quant. Structure-Activity Relationships QSAR and Molecular Modeling], 11th, pp. 31 8, Lausanne, Switzerland. 

Computational neural networks (see Neural Networks in Chemistry) can also be used to generate quantitative models to predict physical properties. In the work de.scribed here. 

Figure 4 Plot of observed vs. calculated aqueous. solubilities for the training set, cross-validation set, and prediction set compounds using a 9-3-1 computational neural network model...

A hands-on experience with the method is possible via the SPINUS web service [48. This service uses a client-server model. The user can draw a molecular structure within the web browser workspace (the client), and send it to a server where the predictions are computed by neural networks. The results are then sent back to the user in a few seconds and visualised with the same web browser. Several operations and different types of technology arc involved in the system ... 

We have already met one tool that can be used to investigate the links that exist among data items. When the features of a pattern, such as the infrared absorption spectrum of a sample, and information about the class to which it belongs, such as the presence in the molecule of a particular functional group, are known, feedforward neural networks can create a computational model that allows the class to be predicted from the spectrum. These networks might be effective tools to predict suitable protective glove material from a knowledge of molecular structure, but they cannot be used if the classes to which samples in the database are unknown because, in that case, a conventional neural network cannot be trained. 

Hemmateenejad, B., Miri, R., Safarpour, M.A., Mehdipour, A.R. Accurate prediction of the blood-brain partitioning of a large set of solutes using ab initio calculations and genetic neural network modeling. J. Comput. Chem. 2006, 27, 1125-35. 

Multivariate models using neural networks, support vector machines and least median squares regression have been used to predict hERG activity [96-98]. These types of models function more as computational black box assays. 

---