search

greenelab / mpmp

Multimodal Pan-cancer Mutation Prediction
BSD 3-Clause "New" or "Revised" License
7 stars 6 forks source link

Update figures with results from extended parameter ranges #53

Closed jjc2718 closed 3 years ago

jjc2718 commented 3 years ago

Closes #44 (finally - these results should be pretty close to what we want to submit at this point).

The main change in this PR is rerunning all of the single-omics experiments with the correct gene sets and larger hyperparameter ranges (see changes to mpmp/config.py), and updating all of the figures with the new results.

The change in hyperparameter ranges caused a fairly substantial change to our results and interpretation: the methylation datasets now perform a lot better, generally on par with gene expression data. We think this is probably because the parameters were optimized at some point in Greg's papers to work well with expression data, so expanding the ranges favored the other data types.

Most of the code changes are related to the figures (either refactoring the plotting code to be a bit simpler, or adding information/details to the figures). Nothing too consequential other than the parameter range updates.

jjc2718 commented 3 years ago

Looks good! Figures 1 and 6 aren't displayed in generate_figures.ipynb, though that may be intentional.

Yep, that's intentional (Figure 1 is generated in BioRender/Inkscape since it's an overview figure/schematic thing, and Figure 6 is just directly generated by another notebook with no panel wrangling required). I'll make this clear in the notebook.