scCRISPR-seq Analysis Recipe

[sreichl]

https://github.com/epigen/MrBiomics/wiki/scCRISPR%E2%80%90seq-Analysis-Recipe

pertpy docs contain nice CROP-seq data e.g., Papalexi 2021

[sreichl]

Definitely include the cross-prediction graph (e.g., at the end) and refer to Peter and my paper to show how it can be used to infer similarities between knockout phenotypes.

[sreichl]

• Download data
• Prepare data for scRNAseq pipeline (ie put in 10X format, especially guides)
• ScRNAseq processing seurat
• Mixscape Seurat
• Unsupervised Analysis
• DEA Seurat
• Crossprediction: adapt to work with PRTB data

[sreichl]

• didn't get pertpy to work
• pivot to SeuratData, which also has the data, but only the processed i.e., post-mixscape =/
  • https://github.com/satijalab/seurat-data
  • Processed data are also available at SeuratData (https://github.com/satijalab/seurat-data) to facilitate access with a single command (InstallData(ds=’thp1.eccite’)).

GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE153056

[sreichl]

Papalexi 2021 Nat Genet paper: https://www.nature.com/articles/s41588-021-00778-2

[sreichl]

• don't do assignment and take their metadata, which includes gRNA and KO assignment for now and continue/move on -> yes
• make issue to provide protospace_thresholds in the config to enable scCRISPR-seq workflows also for MTX format

Unnecessary alternatives

• grna threshold of 1 probably too low as there are mainly non-singlets, although they only report "singlets" -> look at distribution of gRNA counts and select new threshold.
• repeat with data from GEO for the real recipe where Mixscape is still required/useful etc.
  • find(!) & download RAW count data -> at first glance couldn't identify Gene Expression counts, only GDO, ADT, HTO....
• Alternatively, directly download the mudata object from figshare and check if it is also already the Mixscape-processed data: https://figshare.com/ndownloader/files/36509460 see here

[sreichl]

Mixscape Vignette uses the exact same dataset: https://satijalab.org/seurat/articles/mixscape_vignette

[sreichl]

• rerun with all features instead of HVG (rename previous results to _HVG)
• correct for orig.ident i.e., lane and(?) replicate?
• enrichment analysis to connect to F3: all _TA gene sets as a database to check KO phenotypes against via Preranked GSEA? Does that make sense?

[sreichl]

There are only 48 SPI1 cells in the whole dataset, of which 20 are perturbed, i.e., non-escaped. I checked.

[sreichl]

• select plots for wiki
  • plot more enrichment results in the summary than 1 per KO e.g., 3?
• create custom database from CRISPR screen DEA results eg for EpiDrug downstream usage
• write wiki
• update Snakemake report on webserver