Compare functional scores to functional effects inferred from global epistasis model

[Caleb-Carr]

Create a correlation plot for single mutant variant scores and the deconvolved single mutant functional effects inferred from the global epistasis model.

[Caleb-Carr]

@jbloom This notebook compares the functional scores calculated for barcoded variants with single amino-acid mutations to the functional effects of mutations inferred from the global epistasis model. Overall, seems like scores are highly correlated.

[jbloom]

Yes, this looks good. I would suggest adding a few additional things:

• The times_seen parameter you report as a tooltip is a bit confusing as it is the times_seen only for the global epistasis functional effects (as I understand it). Can you report a times_seen for both all barcoded variants and just single amino-acid mutant variants.
• Can you make the same correlation plot conditioning on just mutations that are seen at least twice as single-amino-acid mutant variants? It should probably be higher but with fewer total mutations.
• Can you tabulate how many total mutations are measured using the single amino-acid variant scores versus the global epistasis model? We can use this to explain why were are using the epistasis model: we get many more mutations.

Then after that this should basically be good and we can convert to a supplemental figure.

[Caleb-Carr]

Sounds good! Just one follow up question related to point one above. Does this mean add a tooltip that says for X mutation, it is seen Y times across all barcoded variants and Z times in single mutant barcoded variants?

[jbloom]

Yes

[Caleb-Carr]

@jbloom Updated notebook:

• To show times_seen for both single mutant variants and all variants as a tooltips
• To show correlation subsetted on single mutants seen in >= 1 barcoded variants and >=2 barcoded variants
• To show the number of mutations measured with single mutant variants (7566 mutations). Will use to make table for mutations measured as single mutants and all mutations measured with multi-mutant variants

[jbloom]

Sounds good. For the paper I think we should probably add in the supplement the plot with the >= 2 times seen in single mutants, since that is the comparable threshold to what we use for times_seen for global epistasis fit models.