alexlenail / perturbseq_gi

Code for reproducing "Exploring genetic interaction manifolds constructed from rich single-cell phenotypes"

Perturbseq_GI

These notebooks contain code reproducing the single-cell analyses from:

Norman, T.M., Horlbeck, M.A., Replogle, J.M., Ge, A.Y., Xu, A., Jost, M., Gilbert, L.A., & Weissman, J.S. "Exploring genetic interaction manifolds constructed from rich single-cell phenotypes", Science, 2019.

This repository also contains a version of a library for loading and manipulating Perturb-seq experiments (in the perturbseq subdirectory). A fully self-contained tutorial for using this library can be found in the perturbseq_demo repository, and it may be useful to go through that before attempting to use these notebooks. Finally, this repository also contains a Numpy implementation of the Maxide method for constrained matrix completion (in maxide.py).

In order to use the notebooks, you will need to download the sequencing data from GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE133344. Only the outputs from cellranger are necessary. They should be placed in a directory structure mimicking the output of cellranger (i.e. with an outs folder and appropriate raw_gene_bc_matrices_mex and filtered_gene_bc_matrices_mex subdirectories). The cell_identities.csv files should be placed in the outs folder. (You can go through the tutorial in the perturbseq_demo repository for an example of the expected structure.)

The notebooks are commented but are not "production" software. As such some tinkering will be necessary to get dependencies installed and data into appropriate locations. The appropriate starting point is the notebook GI_generate_populations which does basic loading and normalization of the raw single-cell sequencing data. Most of the notebooks expect to be run from Python 2.7 kernels, though the underlying Perturb-seq library is compatible with Python 3 as well.