Add main figure 3

[jenna-tomkinson]

Add main figure 3

In this PR, the model data was updated that will be used for model evaluation and generated main figure 3 for the manuscript/presentation.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[gwaybio]

Figure nitpicks incoming!

Panel A

• There's no need to specify "all_plates" in the facet (no need to facet at all)
• The font size is a bit small everywhere, particularly the labels above each bar
• Please add context to the grouped bar labels. "Final" doesn't mean much to someone who is unfamiliar with the work. Can you think of a different word to describe this? Similarly, "Shuffled" doesn't mean much either. What is shuffled? Perhaps labels "NF1 genotype\npredictions" and "Negative control" 🤷 ?
• The x axis label is "Data Type", but this isn't specific enough either - maybe no label is the way to go?
• The legend title is "Data Splits", but this is also a bit vague. Perhaps "Machine learning\nmodel data\nsplits"?
• While "rest" and "test" make sense for us internally, it is ok to change for the public-facing visualization. I would go back to "Train" and "Test".
• Note the consistent capitalization throughout my suggestions above :)

Panel B

• Make sure to transfer consistent terminology in Panel B when you update Panel A
• Same comment about the facet too

Panel C

• Make sure to transfer consistent terminology in Panel C when you update Panels A/B (e.g., test/rest)
• Whitespace management probably can be improved. Is there another useful panel you can make for D? Maybe example image? Or I might recommend making C on the same row (not two rows)
• The font size and boxes can prob be bigger (especially font size within boxes)
• Camel case "Predicted Genotype" should be title case ("Predicted genotype")

[jenna-tomkinson]

@gwaybio I have addressed the first round of nit-picks! 😄 Feel free to re-review when possible!

[gwaybio]

LGTM! Here are some final minor nitpicks

• The precision recall curve is not coord_fixed() (or close to having the same x and y axis distance - it should be close to square). You can shrink the height of the top row in the patchwork to make it more square, or, you can swap panels B and C
• The whitespace in the confusion matrix is still a bit off between squares. Specifically, the horizontal distance is about 2x greater than the vertical distance (e.g., in top left, the distance between 885 and 227 is ~2x greater than 885 and 214.)

This plot looks great otherwise, feel free to merge at will!

[jenna-tomkinson]

@gwaybio I have finished the second round of revisions and I will merge now! Thank you for the reviews!