The PMut offer these services:

  • Repository: in the repository you will find proteins for which we have predicted all their possible mutations to be Neutral or Pathological. For now, more than 100,000 proteins have been added to the repository, including all human proteins from UniProtKB. Check this example protein.
  • Analyze mutations: by providing a protein sequence or finding one using our search tool, you can analyse the mutations of your choice using our PMut2017 predictor (or even a predictor created by yourself). Check this example analysis.
  • Batch analysis: you can predict the pathology of mutations on more than one protein. You can provide the proteins' sequences in FASTA format. Check this example batch analysis.
  • Custom predictor: you can train a custom predictor for your later use in new analyses. You must provide a training set with mutations annotated as neutral or disease causing, and the system will generate a new predictor using this data. Check this example predictor.
  • API: use the API to access all the repository data programatically. You can retrieve predictions and features of mutations, information related to a protein, etc.
  • PyMut Python module: this Python module brings all the PMut funtionality to your computer, allowing you to work locally on your mutations and customize the machine learning process. Check the PyMut tutorial to see it in action.


In the following screenshot we can see a page of the repository. Using the Search box we can filter and find any protein we are looking for.

Each protein in the repository has a page where we can find the predicted pathology score for each mutation. We can also find information related to the protein, such as known variants, structural annotations, alignment files and a 3-D visualizer in case a PDB structure is available for the protein.

The chart at the top of the page can be zoomed using mouse scroll, and summarizes the predicted pathology score for all possible mutations.

In the following image we can see and example of a 3-D structure of a protein:

Analyze mutations

By using the Analyze mutations form you can create a custom analysis. In the following screenshot you can see an example analysis results page:

Batch analysis

Just like in the previous case, we can submit a batch analysis, getting a results page such as this one:

Custom predictor

By using the Custom predictor form you can create a custom predictor. In the following screenshot you can see an example predictor page. It includes:

  • Summary: with the computation status and descriptive information.
  • Training set: where you can check the data used to train the predictor.
  • Features: the features computed for each mutation are plotted as historgrams.
  • Evaluation: the predictor precision is evaluated using different methods chosen in the form. The results are provided in a table and with a ROC curve plot.

Predictor summary:

Predictor training set:

Predictor features histograms:

Predictor evaluation: