Skip to contents

Compute the quality of a screen from reference genes

Source: R/reference_genes.R
compute_screen_quality.Rd

Measures how well a screen separates genes known to be essential from genes known to be non-essential, the standard quality check of proliferation (dropout) screens, both CRISPR and shRNA. In a good screen the essential genes are clearly more depleted than the non-essential ones. The reference genes can be the core essential and non-essential genes of Hart et al. (hart_reference_genes) or any other lists.

Usage

compute_screen_quality(
  table,
  essential,
  nonessential,
  score = "Log2FC",
  level = c("gene", "barcode")
)

Arguments

table

A table with one row for each barcode and the columns Gene and score, for example the output of compute_metrics or compute_barcode_statistics.

essential

The names of the essential genes (as in the Gene column).

nonessential

The names of the non-essential genes.

score

The column with the score (default "Log2FC").

level

"gene" (default) to average the scores of the barcodes of each gene, or "barcode" to use each barcode.

Value

A one-row tibble with the columns NEssential and NNonessential (the number of genes or barcodes used), MeanEssential, MeanNonessential, NNMD, CohenD and AUC.

Details

The scores (by default the log2 fold changes) are averaged over the barcodes of each gene (level = "gene") or used barcode by barcode (level = "barcode"). With \(x_E\) the scores of the essential genes and \(x_N\) those of the non-essential genes:

  • NNMD, the null-normalized mean difference (Hart et al. 2015): \((mean(x_E) - mean(x_N)) / sd(x_N)\). More negative is better.

  • Cohen's d: the difference of the means divided by the pooled standard deviation. More negative is better.

  • ROC AUC: the probability that an essential gene has a lower score than a non-essential gene (0.5: no separation, 1: perfect separation).

The reference genes not present in the table are ignored; the number of genes used is reported.

References

Hart T, Chandrashekhar M, Aregger M, et al. (2015). High-resolution CRISPR screens reveal fitness genes and genotype-specific cancer liabilities. Cell 163:1515-1526. doi:10.1016/j.cell.2015.11.015

See also

scale_by_reference, plot_reference_separation, hart_reference_genes

Examples

data(k562_count_table)
data(k562_annotation_table)
data(hart_reference_genes)
samples <- data.frame(
    Sample = c("Plasmid", "T14_A", "T14_B"),
    Day = c("Plasmid", "T14", "T14"), Treatment = "Untreated"
)
object <- create_screenr_object(
    table = k562_count_table, annotation = k562_annotation_table,
    groups = factor(c("Plasmid", "T14", "T14")), sample_table = samples
)
object <- compute_data_table(normalize_data(object))
metrics <- compute_metrics(object,
    control = "Plasmid", treatment = "T14", condition_column = "Day"
)
essential <- hart_reference_genes$Gene[
    hart_reference_genes$Class == "core essential"]
nonessential <- hart_reference_genes$Gene[
    hart_reference_genes$Class == "non-essential"]
compute_screen_quality(metrics, essential, nonessential)
#> # A tibble: 1 × 7
#>   NEssential NNonessential MeanEssential MeanNonessential  NNMD CohenD   AUC
#>        <int>         <int>         <dbl>            <dbl> <dbl>  <dbl> <dbl>
#> 1        660           890         -1.07          -0.0380 -2.39  -1.49 0.838