Correct CRISPR-Cas9 copy-number and proximity bias with CRISPRcleanR

In CRISPR-Cas9 screens, sgRNAs that cut in amplified regions of the genome (or in clusters of nearby targets) deplete the cells whatever the gene they target, producing false essential genes. This function corrects the counts of a ScreenR object with the CRISPRcleanR package (Iorio et al. 2018), which finds these regions by segmenting the sgRNA fold changes along the genome, without needing copy-number data. ScreenR does not implement the correction itself: it prepares the input, calls the CRISPRcleanR functions and returns a new ScreenR object with the corrected counts, ready for normalize_data and the hit-finding functions.

Use it for CRISPR-Cas9 knockout screens only (not for shRNA, CRISPRi or CRISPRa screens), with a library annotation that gives the genomic position of each sgRNA, and on the complete genome-wide library: CRISPRcleanR estimates the bias from runs of neighbouring sgRNAs that target different genes, assuming that most of them have no effect. On a subset of genes enriched in hits (for example only the reference genes, or the hap1_count_table example), whole segments of essential genes look biased and their real signal is removed.

Usage

correct_crispr_bias(
  screenR_Object,
  library_annotation,
  control,
  treatment,
  min_reads = 30,
  min_targeted_genes = 3,
  label = "screen",
  outdir = tempfile("crisprcleanr_"),
  verbose = FALSE
)

Arguments

screenR_Object

A ScreenR object created with create_screenr_object from a CRISPR-Cas9 screen.

library_annotation

The sgRNA library annotation in the format of CRISPRcleanR: a data frame with the sgRNA identifiers (the barcodes of the ScreenR object) as row names and at least the columns GENES, CHRM (chromosome), STARTpos and ENDpos. CRISPRcleanR includes the annotation of several libraries, for example data(KY_Library_v1.0, package = "CRISPRcleanR") or Brunello_Library, AVANA_Library, GeCKO_Library_v2.

control

The name(s) of the control samples (for example the plasmid or T0), as in the columns of the count table.

treatment

The names of the treated (or late time point) samples.

min_reads

sgRNAs with fewer reads than this in the controls (mean) are removed (default 30, the CRISPRcleanR default).

min_targeted_genes

Only segments with sgRNAs targeting at least this number of different genes are corrected (default 3, the CRISPRcleanR default).

label

A name for the screen, used by CRISPRcleanR in its messages and output files.

outdir

The directory where CRISPRcleanR writes its intermediate files (by default a temporary directory).

verbose

Show the progress messages of CRISPRcleanR (default FALSE).

Value

A ScreenR object with the corrected counts of the control and treated samples. S4Vectors::metadata(object)$crisprcleanr is a list with the uncorrected sgRNA log2 fold changes (logFCs), the corrected fold changes (corrected_logFCs) and the segments (segments) returned by CRISPRcleanR.

Details

CRISPRcleanR is not on CRAN or Bioconductor, so it is not installed with ScreenR. Install it with BiocManager::install("francescojm/CRISPRcleanR") (or remotes::install_github("francescojm/CRISPRcleanR")).

The function runs the standard CRISPRcleanR workflow on the given control and treated samples:

1. ccr.NormfoldChanges: removes the sgRNAs with fewer than min_reads reads (mean of the controls), normalizes the counts and computes the sgRNA log2 fold changes;

2. ccr.logFCs2chromPos: sorts the fold changes by genomic position;

3. ccr.GWclean: segments the fold changes and corrects the segments with a biased response;

4. ccr.correctCounts: computes corrected counts from the corrected fold changes.

The returned object contains only the control and treated samples and the sgRNAs kept by CRISPRcleanR. Its counts are the corrected normalized counts of CRISPRcleanR (rounded to integers, so that edgeR can model them). The CRISPRcleanR results are in S4Vectors::metadata(object)$crisprcleanr.

References

Iorio F, Behan FM, Goncalves E, et al. (2018). Unsupervised correction of gene-independent cell responses to CRISPR-Cas9 targeting. BMC Genomics 19:604. doi:10.1186/s12864-018-4989-y

See also

compute_screen_quality to check the screen with reference genes before and after the correction.

Examples

# Needs the CRISPRcleanR package, which is not on CRAN or Bioconductor:
# BiocManager::install("francescojm/CRISPRcleanR")
if (crisprcleanr_available()) {
    ccr <- "CRISPRcleanR"
    # The KY library annotation distributed with CRISPRcleanR
    KY_Library_v1.0 <- get(utils::data("KY_Library_v1.0", package = ccr))
    file <- system.file("extdata", "HT-29_counts.tsv", package = ccr)
    counts <- read.delim(file)
    # A few chromosomes to keep the example fast
    keep <- KY_Library_v1.0[counts$sgRNA, "CHRM"] %in% c("8", "17")
    counts <- counts[keep, ]
    colnames(counts) <- c("Barcode", "Gene", "Plasmid", "R1", "R2", "R3")
    object <- create_screenr_object(
        table = counts[, -2],
        annotation = counts[, c("Barcode", "Gene")],
        groups = factor(c("Plasmid", "Late", "Late", "Late")),
        sample_table = data.frame(
            Sample = c("Plasmid", "R1", "R2", "R3"),
            Day = c("Plasmid", "Late", "Late", "Late"),
            Treatment = "Untreated"
        )
    )
    corrected <- correct_crispr_bias(object,
        library_annotation = KY_Library_v1.0,
        control = "Plasmid", treatment = c("R1", "R2", "R3")
    )
    corrected
}