Added consensus moa analysis files

[AdeboyeML]

@gway I have added the jupyter notebook that contains the code for the analysis on the consensus datasets to get the moa sizes/values per doses(1-7), I have also included the results of the analysis i.e. the heatmap plots and the csv files (folders).

[gwaybio]

this looks great @AdeboyeML - a couple of immediate things to change, before I take a deeper dive.

1. Let's move this entire analysis into a folder called 1.data-exploration.
2. Let's be sure to keep data extensions on csv files. For example, moa_sizes_consensus_datasets/consensus_median_moa_sizes should be moa_sizes_consensus_datasets/consensus_median_moa_sizes.csv.

The reason for 1 is that we want to perform several analyses in this repo, and this is just one component. Each new analysis will belong in an analysis submodule.

[AdeboyeML]

@gwaygenomics Thanks for the observations.

I have created the 1.Data-exploration folder (putting all files and sub-folders into the folder), and also updated the notebook with the null distribution analysis and I have also saved all the dataframe outputs as .csv files.

The .csv files has both the median_scores and p_values dataframes for each consensus dataset per dose (for each of the 231 MOAs with more than one compound).

[gwaybio]

Looks great @AdeboyeML - sorry this has taken me so long to review. I am just about ready to approve.

Two major things:

1. It looks like there is a large number of cells in the notebook that are not executed. Can you either execute the cell or delete the cells? Jupyter notebooks need to be executed sequentially, and if I am viewing a notebook without sequential execution, I have extreme doubts about the data analysis. (not saying that I doubt this one, just in general it is suboptimal practice)
2. Can you add in the figures you generated in https://github.com/broadinstitute/lincs-profiling-comparison/issues/2#issuecomment-723839481? They look great, and are not currently captured in this PR. I would also switch to histograms from density plots (in the seaborn jointplot) since the density tails extend far beyond p < 0 (which doesn't make sense).

Other pull request advice

• I definitely appreciate the scope of this PR. It is not huge. I have reviewed big pull requests in the past and they are not fun! However, the jupyter notebook is quite long. In the future, we should consider including fewer analyses in a single jupyter notebook, and instead number the notebooks if they require sequential execution. (e.g. consensus_moa_sizes_per_dose_analysis.ipynb becomes 0.consensus_moa_sizes_per_dose_analysis.ipynb and one other script to visualize results, or something)
• It is also good practice to include the nbconverted file from the .ipynb files. These nbconverted files can be generated with the following command: jupyter nbconvert --to=script --FilesWriter.build_directory=scripts/nbconverted *.ipynb. The nbconverted files are so much easier to review than .ipynb files.

Advice for myself

• I think I might be needlessly holding you up on performing the next analyses, and I don't want to be! We value good and reproducible code extremely highly, but I need to be aware of when this search for perfection slows things up too much!

[shntnu]

Dropping in with a minor comment: Please add the column name for the first column (moa) to the CSV files e.g. https://raw.githubusercontent.com/broadinstitute/lincs-profiling-comparison/9bc5db8167674e2c8bec5cee3fcc043117acfbf6/1.Data-exploration/moa_sizes_consensus_datasets/median_dmso_null_p_values.csv

[gwaybio]

@AdeboyeML - we should prioritize merging in this contribution early next week. Here is a todo list:

• [ ] Remove unexecuted cells and non-essential comments
• [ ] Make sure all cells are executed in order
• [ ] Make sure all columns have names in all data that is output (nice catch @shntnu)
• [ ] Confirm null distribution (maybe exclude this for now - can you add an issue stub saying this will be added in the future)
• [ ] Split the long notebook into four (or five) different, independent notebooks

I see people forgetting to add column names for indices all the time, and it bugs me! It is also a very easy fix:

data_moa_values = data_moa_values.reset_index().rename({"index": "moa"}, axis="columns")
save_to_csv(data_moa_values, 'moa_sizes_consensus_datasets', 'modz_moa_median_scores.csv')

[AdeboyeML]

@gwaygenomics - thanks for the observations and major points highlighted.

• I have done all necessary corrections to the jupyter notebooks and csv files. I have re-pushed new notebooks and files to the lincs-profiling-comparison repo containing notebooks for different profiling technologies and have also created new independent notebooks for different use cases.

[gwaybio]

once you address these three things:

• [ ] Add citation
• [ ] Remove .ipynb_checkpoints file and add to .gitignore file (example here)
• [ ] Add .py files for corresponding .ipynb files

Go ahead and merge.

Nice work! 🎉

[AdeboyeML]

@gwaygenomics all highlighted points have been addressed and corrected.

[gwaybio]

LGTM 👍 please merge at will

[AdeboyeML]

@gwaygenomics I do not think I have the permission to merge the pull request.

Only those with write access to this repository can merge pull requests.

[gwaybio]

ah, right. Just made you admin. You need to accept in an email, then you should be able to merge. Thanks for pointing this out!