Computes, for each barcode, the log2 fold change between the
treated and the control samples of a given time point (or between two
time points, see condition_column), and its Z-score
and robust Z-score. The result is the input of the Z-score methods
(find_gene_zscore_hit, find_zscore_hit) and
of several plots. More information about the Z-score and other metrics
used in genetic screens can be found in this
paper.
Usage
compute_metrics(
screenR_Object,
control,
treatment,
day = NULL,
pseudocount = 0.5,
condition_column = c("Treatment", "Day")
)Arguments
- screenR_Object
The ScreenR object obtained using
create_screenr_object.- control
The value(s) of the condition_column column of the data table (see
get_sample_table) that identify the control samples.- treatment
The value(s) of the condition_column column of the data table that identify the treated samples.
- day
The day (time point) to use, as in the Day column of the data table. Required with condition_column = "Treatment"; not used with condition_column = "Day".
- pseudocount
A small value (in counts per million) added to the treated and control mean frequencies before computing the log2 fold change, so that barcodes with zero reads get a finite fold change: Log2FC = log2((Treated + pseudocount) / (Control + pseudocount)). Default 0.5. Barcodes with zero reads in both treated and control are removed. The value is on the scale of the normalized counts: per million reads of the effective library size of
normalize_data(all the reads for "cpm", all the reads times the TMM factor for "tmm", the reads of the control barcodes for "control"). It is a display transformation that keeps the fold changes finite; it does not change the normalization.- condition_column
The column of the data table that tells the control from the treated samples: "Treatment" (default) to compare a treatment with a control at the time point day, or "Day" to compare two time points, for example a late time point with T0 in a dropout screen without treatment (control = "T0", treatment = "T18").
Value
A tibble with one row for each barcode and the columns Barcode, Gene, Treated and Control (the mean normalized counts), Log2FC, Zscore, ZscoreRobust, Day and Treatment. With condition_column = "Day", Day is the time point of the treated samples and Treatment their treatment.
Details
The function uses the normalized counts (see
get_data_table), so run normalize_data
first. For each barcode the normalized counts of the treated and of the
control samples of the given day are averaged, then
Log2FC = log2((Treated + pseudocount) / (Control + pseudocount)),
Zscore = (Log2FC - mean(Log2FC)) / sd(Log2FC) and
ZscoreRobust = (Log2FC - median(Log2FC)) / mad(Log2FC), where the mean,
standard deviation, median and median absolute deviation are computed
over all the barcodes. Since the replicates are averaged, their
variability is not used: see compute_barcode_statistics
for a replicate-aware alternative.
Examples
object <- get0("object", envir = asNamespace("ScreenR"))
metrics <- compute_metrics(object,
control = "Time3",
treatment = "TRT", day = "Time3"
)
head(metrics)
#> # A tibble: 6 × 9
#> Barcode Gene Treated Control Log2FC Zscore ZscoreRobust Day Treatment
#> <fct> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <chr> <chr>
#> 1 10208 Gene_380 372. 382. -0.0390 -0.0962 -0.138 Time3 TRT
#> 2 10502 Gene_380 32.2 27.9 0.205 0.625 0.857 Time3 TRT
#> 3 10796 Gene_380 140. 123. 0.190 0.579 0.794 Time3 TRT
#> 4 11089 Gene_380 378. 365. 0.0510 0.170 0.229 Time3 TRT
#> 5 11382 Gene_380 48.6 66.9 -0.456 -1.33 -1.84 Time3 TRT
#> 6 11675 Gene_380 691. 829. -0.263 -0.759 -1.05 Time3 TRT
# Two time points of the same samples, as in a dropout screen
metrics_time <- compute_metrics(object,
control = "Time3", treatment = "Time4", condition_column = "Day"
)
head(metrics_time)
#> # A tibble: 6 × 9
#> Barcode Gene Treated Control Log2FC Zscore ZscoreRobust Day Treatment
#> <fct> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <chr> <chr>
#> 1 10208 Gene_380 375. 377. -0.00695 0.185 0.0758 Time4 TRT/Time4
#> 2 10502 Gene_380 32.4 30.1 0.108 0.454 0.421 Time4 TRT/Time4
#> 3 10796 Gene_380 46.7 131. -1.48 -3.28 -4.35 Time4 TRT/Time4
#> 4 11089 Gene_380 307. 372. -0.278 -0.450 -0.737 Time4 TRT/Time4
#> 5 11382 Gene_380 34.4 57.8 -0.738 -1.53 -2.12 Time4 TRT/Time4
#> 6 11675 Gene_380 886. 760. 0.221 0.720 0.761 Time4 TRT/Time4
