Construction of PLS models Using Discovery Studio

Introductions

Discovery Studio is a broad range of computational tools for drug discovery and design. one of the most important features of Discovery Studio is its ability to construct partial least squares (PLS) models that can be used for 3D-QSAR analysis. The process of constructing PLS models involves several steps, including data preparation, model construction, and model validation. In b this tutorial, we will explore each of these steps in detail and discuss how they contribute to the overall success of building PLS models in Discovery Studio.

Tutorials

  • Model Construction

Import the training set molecules

In the file browser, click Samples>Tutorials>QSAR, and double-click trainingset.sd to import the 14 molecules of the training set required for this tutorial.

Import the test set molecules

In the file browser, click Samples>Tutorials>QSAR, and double-click testset.sd to import 8 test set molecules.

Both training set molecules and test set molecules are stacked beforehand, and can generally be stacked directly based on the common backbone of small molecules.

Construct the model

Click Small Molecules>Create QSAR Model>Create 3D QSAR Model in the toolbar, and set the running parameters in the pop-up dialog box.

Input Ligands is set to trainingset:All

Input Test Ligands is set to testset:All

Activity Property is set to arNCTRlogRBA

Enter GridBasedModel in the Model Name field

The rest of the parameters are set to default, click Run to run the task.

  • Model Validation

The final step in constructing a PLS model is to validate the model to ensure that it is accurate and reliable. This involves testing the model on a set of molecules not included in the original dataset (called the test set) and comparing the predicted activity values with the actual bioactivity values.

Reliability evaluation

In the Report interface, click View Results, and click Yes in the pop-up window.

The horizontal coordinate is the experimental activity value of the molecule in the training set, and the vertical coordinate is the predicted value.

Checking the grid point diagram

The grid plot provides some drug design information based on the static and stereoscopic fields, based on which we can modify the design of small molecules to find more active drug molecules. The red area indicates that the more negatively charged the substituent in the region is, the better the activity of the compound; the blue area indicates that the more positively charged the substituent in the region is, the better the activity of the compound.

Conclusion

Building PLS models in Discovery Studio is a powerful tool for drug discovery and design. By carefully preparing data, selecting appropriate descriptors and validating models, researchers can develop accurate and reliable models for predicting the biological activity of new molecules.

Related Services

3D-QSAR Service

* For Research Use Only.
Inquiry