# Third-party CRISPR Guide Assignment Toolkit
[Crispat](https://academic.oup.com/bioinformatics/article/40/9/btae535/7750392) is a peer-reviewed CRISPR guide calling package that conveniently bundles many popular gRNA thresholding approaches, enabling users to compare the performance of multiple approaches side-by-side.
Here, we implement one of these approaches, the Gaussian-Gaussian Mixture Model, which we've found performs best with the Illumina's Single-Cell CRISPR Prep kit.
[Instructions for how to download and install](https://github.com/aamayzhang/crispat_iscp/tree/main) our optimized fork of crispat can be found[ here](https://github.com/aamayzhang/crispat_iscp/tree/main).
In brief, the model does the following:
* Read in DRAGEN filtered matrix files from an DRAGEN output directory.
* Creates and reads in a .h5ad file (via scanpy).
* Performs 2-component Gaussian-Gaussian mixture modeling on the gRNA counts.
* Outputs gRNA assignments for each cell barcode along with plots showing the fit of the model to the raw counts.
A [step-by-step tutorial](https://github.com/aamayzhang/crispat_iscp/blob/main/tutorial/iscp_demo.py) and a [small test dataset](https://github.com/aamayzhang/crispat_iscp/tree/main/example_data) can be found in the repo.
You should expect the plots of the models fit to nonzero count data to look similar to this:
The output file *assignments.csv* provides a full list of all cell barcodes containing a guide that exceeded the threshold determined by the gaussian mixture model. The outputs should look as follows:
`cell`: the cell barcode sequence.
`gRNA` : the gRNA ID that was assigned to the cell.
`read_counts`: the number of read counts observed for each cell-gRNA assignment.
