file of predictors (or notebook that generates such a file)

[jbloom]

@marlinfiggins, this relates to our discussion of generating predictors for the model.

I propose to write a notebook that will collate various predictors. I propose it will output two files.

The first will be a mutation-effect file, and will have the following columns:

• study
• phenotype
• parental amino acid
• site
• mutant amino acid
• effect

The second will be a per-Pango-clade file, and will have the following columns:

• study
• phenotype
• Pango clade
• predicted phenotype
• mutations lacking data (are there mutations in the Pango clade that lack data)

Does this sound good? Is there any other information that would be needed.

For the per-Pango-clade file, I would propose that you can pass a parameter that specifies the reference clade, and predicted phenotypes are calculated with respect to that reference clade.

[marlinfiggins]

This seems to be exactly what we need. In the future, I think it would be useful to have a separate script for creating the per-Pango-clade file within this workflow as well. I imagine that we would to update as new clades are added, but I can do this step later based on your notebook.

[jbloom]

@marlinfiggins, I've just created a GitHub repo that should do this: https://github.com/jbloomlab/SARS2-spike-predictor-phenos

My suggestion would be that you make this a git submodule in your pipeline, perhaps?

Look it over and raise issues if you can't understand it or it isn't doing the correct thing. If it works for your purpose, feel free to close this issue. I am also going to post to Slack channel re this.