Minor tweaks to Sup Figure 8

[gwaybio]

I also needed to adjust the data processing to account for updated input data types.

Supplementary Figure 8

Supplementary Figure 8. Evaluating multiclass predictions of single-cell phenotypes using AreaShape and Zernike CellProfiler feature subsets.

(A) Confusion matrices comparing models trained on real data vs. shuffled data with CellProfiler AreaShape and Zernike feature data subsets only. The number in each box represents the total count and the color represents the ratio of count over ground truth label. All data show test set performance. (B) Precision recall curves for all 15 phenotypes. The shuffled baseline models (dashed line) performed very poorly for all phenotypic classes. (C) F1 scores for test set predictions for 15 phenotypes and overall performance across all five feature sets.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[jenna-tomkinson]

LGTM! I made a few minor comments regarding the supp figure for you to address.

[jenna-tomkinson]

I have a few overall (probably more nit-pick) comments:

1. I think the font size could be increased just a bit, but totally optional. I think panels B and C can be a bit harder to read.
2. In Panel B, the different color greens for Zernike are a bit hard to distinguish, in my opinion, especially when there is overlap as compared to the blue hues in the AreaShape. I would recommend making the green hues more distinguishable.
3. In Panel C, a very nitpicky comment is that the color palette clashes a bit given the new AreaShape and Zernike colors but they are still distinguishable. So it might be good to find less clashy colors or leave it, both work.

[gwaybio]

Thanks for these comments! Here are my replies:

1. While on the small size, fonts look ok to me. We'll have to figure out a way to increase font size if a reviewer requests this, but otherwise it is ok. I actually think the axis labels of panel A are the most difficult to read, so I have updated these slightly in the next commit!
2. Thanks for calling this out! I've updated the color scheme - I think this is much improved :)
3. In addressing point 2, I think I've achieved an answer for point 3.

Merging this in after adding new commits, thanks again

[jenna-tomkinson]

Just curious, how does Rscript formatting work, and should there be two empty endlines at the bottom of the script?

[gwaybio]

nope! I have removed an extra empty cell in the .ipynb file. thanks for noting this!

[gwaybio]

Updated figure