Peptide QSAR Panel

Using the Peptide QSAR Panel

This panel can be used to build and apply a model that relates properties of peptide sequences to an observed property. The independent variables in these models are sets of amino acid properties taken from the literature [7–9], and do not need to be explicitly provided. Only the observed property is needed, for the training set and the test set of the model. Model building builds multiple models for increasing numbers of partial least squares factors.

To set up a model:

1. Click Load Sequences and Observables.

   In the panel that opens, you can choose the source of the sequences and the observables. The sequences can come from peptide structures in the project, from a Fasta file, or from a CSV file. The observables can be taken from the project, if the sequence came from a project, or loaded from a CSV file, entered manually after importing the sequences, or left unspecified if the peptide is one for which you want a prediction.

   When you have chosen the source of sequences and observables, click OK. The panel closes, and the sequences table is loaded with the sequences. Any existing sequences are replaced. You should therefore ensure that all the sequences you want to use can be loaded from the same source.

2. Select Build a new model, choose the method from the QSAR method option menu, and choose the descriptor type from the Peptide descriptor type option menu.

3. Choose the training set and the test set. There are three options:

   • You can choose the sets explicitly. Select Apportion training and test sets based on "Training/Test" column. Select the rows for one of the sets in the table, choose the set from the Set selected rows as option menu, and click Update. Do the same for the other set.
   • You can use all the sequences, and assign the training and test sets randomly. Select Use all rows in the table, and set the percentage for the test set in the Randomly select text box.
   • You can use a subset of the sequences, and assign the training and test sets randomly. Select Use only rows marked as "either", and set the percentage for the test set in the Randomly select text box. Select the rows to use in the table, choose Either training or test from the Set selected rows as option menu, and click Update.

4. Click Build.

   The model might take a minute or so to build, and then the results are displayed in the Results tab.

5. If you want to apply this model to other sequences, you must save it first. Click Export Model to save the model to a file.

To apply a model:

1. Load the sequences you want to apply the model to, if you do not already have the sequences loaded. (See above).

2. Choose the QSAR method for the model you want to apply.

3. Select Apply a model.

4. Click Browse to navigate to the model file and open it.

5. Choose an option for the table rows to apply the model to.

   You can apply it to all rows, or to rows that are marked in the table as training, test, or either, or to rows that are not marked.

6. Click Apply.

   The model is applied and the results are displayed in the Results tab.