Code for identifying cohort-specific oncogenic dependencies and cohort-insensitive pan-essentials using DepMap + a cohort of CRISPR screens analyzed with MAGeCk.
Main analysis is in the Juypter Notebook. R script (for the log-likelihood ratio testing) is taken from P. Montgomery of the DepMap Consortium, available on GitHub. I've included my results here.
Note: some of the NormLRT values returned are probably erroneous, likely related to bad/failed MLE when estimating skewed t-dist parameters, but I haven't gotten around to examining the DepMap R code closely enough to fix this. The R script is run in my Notebook indirectly via Python/rpy2. If the pre-calculated LRT (available in this repo) file is included, the notebook will use that instead as estimating parameters for ~18k genes is slow.
DepMap data should be downloaded from the DepMap data portal.
You will want:
- CRISPR_gene_effect.csv
- CRISPR_common_essentials.csv
- Achilles_gene_effect.csv
- Achilles_common_essentials.csv
I included DepMap_Selective_Genes.csv which is derived from the relevant column of the table available for here (link directly initiates download), targeting only the combined dataset rows.