Input mapper 2.0 features software#

Model improvement: Another possible reason for bad model performance may be improper feature choice, e.g. The researcher might as well discover that the misclassifications are outliers or that more training data is required.

Investigating possible reasons for the erroneous predictions might aid in detecting errors in the training data, like mis-measured endpoint values. Visually examining (Q)SAR model validation results can aid in understanding the model itself as well as the modeled data, and can furthermore yield the following benefits:Дata curation:It is important to inspect (groups of) misclassified compounds (in case of classification), or compounds with high prediction error (regression). It should be regarded as complementary to the standard statistical validation. The presented approach does not investigate how predictions are made by each model, but rather allows the comparison of actual and predicted activity values in the feature space (see Figure 1).

In this paper, we propose a model-independent visual analysis of validation results employing the 3D viewer CheS-Mapper. In addition, these complex models are often more predictive than intuitive and simpler models. In contrast, many other models can rather be seen as black boxes, like artificial neural networks or support vector machines. Therefore, several model-dependent visualization tools exist. the size of the decision tree is reasonably small). The predictions of these models are easy to comprehend, as long as the number of features that are employed for predictions is not too large (e.g. Some models induced by machine learning approaches, however, are relatively easy to understand, like decision trees, rule learners or nearest neighbor models. One reason for this might be that most (Q)SAR models are the results of applying statistical machine learning approaches to chemical datasets and the resulting models are sometimes opaque and it is commonly not an easy task to extract the reasoning behind a prediction. There is however, to the best of the authorsвЂ™ knowledge, no visualization method yet that incorporates (Q)SAR predictions and validation results. The predictive performance of (Q)SAR models can then be evaluated with numerous statistical validation techniques. These correlations are employed by (Q)SAR models to predict the activity of unseen compounds. Many approaches are being developed that help to understand existing correlations between the structure of chemical compounds, their physico-chemical properties, and biological or toxic effects. Visualization of (quantitative) structure-activity relationship ((Q)SAR) information in chemical datasets is a very active field of research in cheminformatics. Graphical abstractД¬omparing actual and predicted activity values with CheS-Mapper.

Input mapper 2.0 features software#

The CheS-Mapper software is freely available at. Moreover, the researcher can use CheS-Mapper to inspect how the (Q)SAR model predicts activity cliffs. It reveals, if the endpoint is modeled too specific or too generic and highlights common properties of misclassified compounds. Visual validation with CheS-Mapper enables analyzing (Q)SAR information in the data and indicates how this information is employed by the (Q)SAR model. The approach is generic: It is model-independent and can handle physico-chemical and structural input features as well as quantitative and qualitative endpoints. The tool enables the comparison of model predictions to the actual activity in feature space. The present work describes the new functionalities in CheS-Mapper 2.0, that facilitate the analysis of (Q)SAR information and allows the visual validation of (Q)SAR models. The approach applies the 3D viewer CheS-Mapper, an open-source application for the exploration of small molecules in virtual 3D space.

We propose visual validation, as an approach for the graphical inspection of (Q)SAR model validation results. Even though, a number of visualization tools for analyzing (Q)SAR information in small molecule datasets exist, integrated visualization methods that allow the investigation of model validation results are still lacking. However, classical validation does not support the user in better understanding the properties of the model or the underlying data. Sound statistical validation is important to evaluate and compare the overall performance of (Q)SAR models.