Model applicability domain

PFTA Non Positive Compound is out of model Applicability Domain No alerts for carcinogenic activity Positive SAs for the micronucleus assay... 

The evaluation for aquatic toxicity on daphnids and fish is reported in Tables 12 and 13. Bold values indicate that compounds are out of the model applicability domain (ECOSAR) or that the prediction is not reliable. ECOSAR and ToxSuite are able to predict all the selected compounds while T.E.S.T. fails in prediction for the daphnia toxicity of perfluorinated compounds (PFOS and PFOA). Tables 12 and 13 include also a limited number of experimental results provided by the model training dataset (some data are extracted from USEPA Ecotox database). Predicted results are in agreement for five compounds only (2, 3, 5, 13 and 14) for both endpoints while the predictions for the other compounds are highly variable. 

Tropsha, A., Golbraikh, A. (2007) Predictive QSAR modeling workflow, model applicability domains, and virtual screening. Curr PharmaceutDesign 13, 3494-3504. 

One of the most important problems in QSAR analysis is establishing the domain of applicability for each model. In the absence of the applicability domain restriction, each model can formally predict the activity of any compound, even with a completely different structure from those included in the training set. Thus, the absence of the model applicability domain as a mandatory component of any QSAR model would lead to the unjustified extrapolation of the model in the chemistry space and, as a result, a high likelihood of inaccurate predictions. In our research we have always paid particular attention to this issue (12, 20-27). A good overview of commonly used applicability domain definitions can be found in reference (28). 

Prediction of bioactivity endpoints by means of the MFTA models. Post-filtering of structures that fall outside the model applicability domain and have unreliable predicted values. 

The failure of models to yield accurate predictions is a consequence of either experimental design or differences in the chemical spaces used to develop and test the models. The identification of model Applicability Domain (AD) can differentiate reliable from non-reliable predictions. There are, indeed, many different approaches to determine AD. These methods can be classified in two... 

MLR see multiple linear regression MM see molecular mechanics model applicability domain, 3, 68, 74 Model scope, 2,155... 

In the following, we will denote the correlation between experiment and prediction for any data set which was not part of the training set by For such a data set, it has to be checked whether the model, in principle, is able to predict these compounds. Despite its paramount significance, the model applicability domain has been addressed in only few QSAR studies [75,78,81,85,91]. 

Following the concept of model applicability domains, a validation of the applicability domain of a global model is strictly advisable for a reasonably sized compound set >10. Within this series, an excellent correlation of the predicted metabolic lability to the experimental data was observed with a correlation coefficient of r =0.80 (Figure 11.4). Thus, the global model for human microsomal lability is applicable for the optimization of this chemical series. Metabolically stable compounds were finally obtained by introduction of a benzoylvaline moiety yielding compound 2 with an ICj of 30 nM in the enzymatic test (metabolic lability of 1%, log D=3.0 atpH=7.4, PSA= 132.45A"). 

Baskin II, Kireeva N, Vamek A (2010) The One-class classification approach to data description and to models applicability domain. Mol Inf 29(8 9) 581 587. doi 10.1002/ minf201000063... 

A QSAR procedure tries to minimize the error of prediction, for example, in the form of the sum of squares between predicted and observed activities. The process of QSAR model development can be divided into three parts data preparation, data analysis, and model validation (O Fig. 37-1). Model validation should include establishment of model applicability domain (AD). Recently, the European Organization for Economic Co-operation and Development (OECD) developed a set of principles for the development and validation of QSAR models, which, in particular, requires appropriate measures of goodness-of-fit, robustness, and predictivity (Organisation 2008). The OECD guidance document especially emphasizes that QSAR models should be rigorously validated using external sets of compounds that were not used in the model development. 

---