Cleanup of Fig 2

[ErinWeisbart]

Outstanding questions for @MerajRamezani:

• [x] Where was the CORUM input downloaded from?
• [x] Please describe len(hit_set), cluster_count, and len(hit_cluster_list_list) in Fig 2C.
• [x] Figure 2F doesn't explain how you determine the number of components that represent 70% variation, it just has 140 hardcoded in. How did you get to the number 140?
• [x] Can you explain why you have used GO terms everywhere in this project for grouping genes except you used KEGG for mTOR_signaling_core in 2H/I

Changes to note: Using the updated CCLE data has a small effect on our hit calling as the control groups are slightly changed. (The lengths of the zero_tpm_list and expressed_gene_list have changed from 3162/17230 to 3067/17325). This has a ripple effect causing minor changes to many numbers/figures in the notebook. Visible changes include: Fig 2A/B: The number of compartment-specific and whole cell hits has changed from 1336/2309 to 1340/2321. This also causes a minor shift in the Number of Hit Genes Called by Compartment pie chart. Fig 2C: The gene pair correlation dictionary has more entries from 6641190 to 6699630. Fig 2E: Shift in compartment ratios

[MerajRamezani]

@ErinWeisbart The Corum data set is from here : http://mips.helmholtz-muenchen.de/corum/#download

I have been using the Corum 3.0( corum_2018_09_03) the 2018 release. They have actually released a new version 4.1, should I update the data?

[ErinWeisbart]

Thanks Meraj. I'll update it to 4.1 in the cleanup I'm doing now. If we do decide to incorporate the PCA into the CORUM analysis as you've described in https://github.com/broadinstitute/pooled-cell-painting-data-analysis/issues/112#issuecomment-1353639813, please wait to update the code until I've merged my cleanup into main.

[ErinWeisbart]

@MerajRamezani I have a working first draft of this notebook. Please address the outstanding questions I have listed above as I would like to address them before merging this in. (You are welcome to provide additional feedback at this point.)

[ErinWeisbart]

From conversation with Meraj and some new code he sent me:

• len(hit_set), cluster_count, and len(hit_cluster_list_list) in Fig 2C are now described
• PCA for CORUM analysis is added
• To determine PCA cutoff Meraj had looked at the PCA and manually chosen a number. I added code for determining it empirically.

I'm going to go ahead and merge this in as the majority of the cleanup is done and will move open questions/known cleanup needs to #12, noting that it is currently an outstanding question which version of the CCLE data we will use in the final version.