To build a classification model on continuous data (e.g. pIC50 data), an interval defining the desired value range is needed. QSARs are managed in the "QSAR" tab of Makya. Please see QSAR Set Up for a brief video explanation.
To define your TPP, click the New QSAR button.
You will then be prompted to select the Makya dataset on which to train your QSAR model.
Once a dataset is selected, the target objectives it contains are shown.
Objectives can then be configured to represent the TPP, using the sliders and/or text fields to set up lower and/or upper bounds for each of the desired objectives. When only a lower or an upper bound selected, Iktos continous classification model is applied.
A QSAR model has to receive at least one objective, and can be trained simultaneously on multiple objectives (multiclass classification) from the same dataset. In the example above, the full TPP is:
- pKi IC50 Pi3K ≥ 7
- pKi IC50 mTor ≥ 8.5
- Caco2 permeability ≤ 0.9
When multiple objectives are selected, a risk matrix is displayed in order to signal any risk of anti-correlation that would impact multi-parametric optimization.
For each objective,
- the objective can be removed from the TPP by clicking on the red cross;
- the molecules matching the selected range can be seen by clicking on the eye icon;
- the number of molecules matching or not this objective (molecules in / out) is displayed.
Molecules matching all the objectives simultaneously can be seen by clicking on the eye icon in the blue banner.
When the set-up is done, simply click on the Train QSAR button to start training your model.