Tools for sgRNA calling in direct capture Perturb-seq data
This repository contains scripts and notebooks used in "Scalable single-cell CRISPR screens by direct guide RNA capture and targeted library enrichment", Replogle JM et al., Nature Biotechnology 2020.
At present, these are simply presented as used to process a single lane of 10x data in the manuscript, with notebooks serving as examples.
The inputs to mixed_model_guide_calling.ipynb are:
(1) a table where each row represents a read aligned to a unique guide identity, cellranger-corrected cell barcode, and cellranger-corrected UMI
(2) the valid cell barcodes as determined by cellranger based on the scRNA-seq data
These datasets are merged and then collapsed to a table of guide UMI counts for each cell.
Then, for each guide, we fit a Poisson-Gaussian mixed model to determine thresholds for sgRNA positive and negative cells.
The output table "cell_identities.csv" contains for a row for every positive sgRNA - cell call. Cells containing multiple guides will be represented in multiple rows. This table is formatted to serve as input to Tom Norman's https://github.com/thomasmaxwellnorman/perturbseq_demo.