kundajelab / coessentiality

Overview

Companion to "A genome-wide almanac of co-essential modules assigns function to uncharacterized genes".

Contains code to generate co-essential gene pairs, co-essential modules, and modules with cancer type-specific dependencies. Coming soon: code to generate the two-dimensional layout (Fig. 1C).

For the web tool associated with the paper, see coessentiality.net. If you would just like the final 17634 x 17634 matrix of p-values, you can download it here. A corresponding 17634 x 17634 matrix with the sign of each correlation (1 = positive, -1 = negative) is downloadable here. The list of the 17634 genes that form the rows and columns of these matrices can be found here. These matrices are in NumPy's npy format and can be loaded with np.load in Python, the RcppCNPy library in R, and the cnpy library in C/C++.

2D coordinates of the final gene layout can be downloaded here.

Code files

1. gene_pairs.py: generates co-essential gene pairs.
2. modules.py: generates co-essential modules using the gene pairs from #1.
3. cancer_type_dependencies.py: enumerates modules with cancer type-specific dependencies using the gene pairs and modules from #1 and #2.
4. load_screens.py: loads and bias-corrects CRISPR screens. Used by #1 and #3.
5. generate_layout.py: generates gene network for visualization, from the GLS p-value matrix and the gene modules from #2.

Required external files

1. gene_effect.csv: CRISPR screens from the "DepMap Public 18Q3" release at https://depmap.org/portal/download/all/. Required for gene_pairs.py and cancer_type_dependencies.py.
2. sample_info.csv: metadata for the cell lines in gene_effect.csv. Required for gene_pairs.py and cancer_type_dependencies.py.
3. cluster_one-1.0.jar: Java executable for ClusterONE. Required for modules.py.

If using a newer release of DepMap, #1 and #2 can be obtained from the DepMap website.