Statistical external validation and consensus modeling: A QSPR case study for Koc prediction

Gramatica, Paola; Giani, Elisa; Papa, Ester

doi:10.1016/j.jmgm.2006.06.005

Cited by 236 publications

(194 citation statements)

References 41 publications

(90 reference statements)

Supporting

Mentioning

190

Contrasting

Order By: Relevance

“…That is why; we do not consider them as outliers. In closing, the model can be used with high accuracy in this applicability domain, 25 and all these data confirm the stability of our model. In the case of D&D model, 3 this kind of analysis was not reported.…”

Section: Model Application Domainsupporting

confidence: 78%

Computational chemistry study of 3D‐structure‐function relationships for enzymes based on Markov models for protein electrostatic, HINT, and van der Waals potentials

et al. 2008

View full text Add to dashboard Cite

In a significant work, Dobson and Doig (J Mol Biol 2003, 330, 771) illustrated protein prediction as enzymatic or not from spatial structure without resorting to alignments. They used 52 protein features and a nonlinear support vector machine model to classify more than 1000 proteins collected from the PDB with a 77% overall accuracy. The most useful features were: the secondary-structure content, the amino acid frequencies, the number of disulphide bonds, and the largest cleft size. Working on the same dataset used by D&D, in this article we reported a good and simple model, based on the Markov chain models (MCM), to classify protein 3D structures as enzymatic or not, taking into consideration the spatial structure without resorting to alignments. Here we define, for the first time, a general MCM to calculate the electrostatic potential, molecular vibrations, van der Waals (vdw) interactions, and hydrophobic interactions (HINT) and use them in comparative studies of potential fields and/or protein function prediction. The dataset is composed of 1371 proteins divided into 689 enzymes and 682 nonenzymes, all proteins were collected from the PDB. The best model we found was a linear model carried out with the linear discriminant analysis; it was able to classify 74.18% of the proteins using only two electrostatic potentials. In the work described here, we define 3D-HINT potentials (mu(k)) and use them for the first time to derive a classifier for protein enzymes. We analyzed ROC curves, domain of applicability, parametric assumptions, desirability maps, and also tested other nonlinear artificial neural network models which did not improve the linear model. In closing, this MCM allows a fast calculation and comparison of different potentials deriving into accurate protein 3D structure-function relationships, notably simpler than the previous.

show abstract

Section: Model Application Domainsupporting

confidence: 78%

Computational chemistry study of 3D‐structure‐function relationships for enzymes based on Markov models for protein electrostatic, HINT, and van der Waals potentials

et al. 2008

View full text Add to dashboard Cite

show abstract

“…Golbraikh and Tropsha [22] have suggested for acceptance of QSAR models a set of criteria, which can be tested during external validation of models. Though cross-validation has been widely criticized in the QSAR literature [22,24,31], it uses all available data for model development and validation, and thus makes the model more reliable simply due to the use of larger number of compounds than the case where a splitting of the data set is performed for external validation purpose. This issue is more important in the case of a small data set, as a significant amount of information is lost due to the omission of some compounds from the training set for external validation.…”

Section: Introductionmentioning

confidence: 99%

On further application of r as a metric for validation of QSAR models

Mitra

Roy

Kar

et al. 2009

Journal of Chemometrics

View full text Add to dashboard Cite

Validation is a crucial aspect for quantitative structure-activity relationship (QSAR) model development. External validation is considered, in general, as the most conclusive proof of predictive capacity of a QSAR model. In the absence of truly external data set, external validation is usually performed on test set compounds, which are members of the original data set but not used in model development exercise. In the case of small data sets, QSAR researchers experience problem in model development due to the fact that the developed models may be less reliable on account of the small number of training set compounds and such models may also show poor external predictability because the models may not have captured all necessary features required for the particular structure-activity relationships. The present paper attempts to show that 'true r 2 m (LOO) ' statistic calculated based on the model derived from the undivided data set with application of variable selection strategy at each cycle of leave-one-out (LOO) validation may reflect external validation characteristics of the developed model thus obviating the requirement of splitting of the data set into training and test sets. This approach may be helpful in the case of small data sets as it uses all available data for model development and validation thus making the resulting model more reliable.

show abstract

“…The QSPR/QSAR models now correlate chemical structure to a wide variety of physical, chemical, biological (including biomedical, toxicological, ecotoxicological) and technological properties. [13][14][15][16][17] QSPR/QSAR models are obtained on the basis of the correlation between the experimental values of the property/activity and descriptors reflecting the molecular structure of the compounds. To obtain a significant correlation, it is crucial that appropriate descriptors be employed.…”

Section: Introductionmentioning

confidence: 99%

Prediction of Melting Point for Drug-like Compounds Using Principal Component-Genetic Algorithm-Artificial Neural Network

Habibi‐Yangjeh

Pourbasheer²,

Danandeh-Jenagharad³

2008

Bulletin of the Korean Chemical Society

View full text Add to dashboard Cite

Principal component-genetic algorithm-multiparameter linear regression (PC-GA-MLR) and principal component-genetic algorithm-artificial neural network (PC-GA-ANN) models were applied for prediction of melting point for 323 drug-like compounds. A large number of theoretical descriptors were calculated for each compound. The first 234 principal components (PC's) were found to explain more than 99.9% of variances in the original data matrix. From the pool of these PC's, the genetic algorithm was employed for selection of the best set of extracted PC's for PC-MLR and PC-ANN models. The models were generated using fifteen PC's as variables. For evaluation of the predictive power of the models, melting points of 64 compounds in the prediction set were calculated. Root-mean square errors (RMSE) for PC-GA-MLR and PC-GA-ANN models are 48.18 and 12.77 ºC, respectively. Comparison of the results obtained by the models reveals superiority of the PC-GA-ANN relative to the PC-GA-MLR and the recently proposed models (RMSE = 40.7 ºC). The improvements are due to the fact that the melting point of the compounds demonstrates non-linear correlations with the principal components.

show abstract

Statistical external validation and consensus modeling: A QSPR case study for Koc prediction

Cited by 236 publications

References 41 publications

Computational chemistry study of 3D‐structure‐function relationships for enzymes based on Markov models for protein electrostatic, HINT, and van der Waals potentials

Computational chemistry study of 3D‐structure‐function relationships for enzymes based on Markov models for protein electrostatic, HINT, and van der Waals potentials

On further application of r as a metric for validation of QSAR models

Prediction of Melting Point for Drug-like Compounds Using Principal Component-Genetic Algorithm-Artificial Neural Network

Contact Info

Product

Resources

About