ct.generateResults: Calculate results of a crispr screen from a contrast

View source: R/GenePvalsFromEset.R

ct.generateResultsR Documentation

Calculate results of a crispr screen from a contrast

Description

This is a wrapper function that enables direct generation of target-level p-values from a crispr screen.

Usage

ct.generateResults(
  fit,
  annotation,
  RRAalphaCutoff = 0.1,
  permutations = 1000,
  contrast.term = NULL,
  scoring = c("combined", "pvalue", "fc"),
  alt.annotation = NULL,
  permutation.seed = NULL
)

Arguments

fit

An object of class MArrayLM containing, at minimum, a t slot with t-statistics from the comparison, a df.residual slot with the corresponding residuals fo the model fits, and an Amean slot with the respective mean abundances.

annotation

An annotation file for the experiment. gRNAs are annotated by row, and must minimally contain columns geneSymbol and geneID.

RRAalphaCutoff

A cutoff to use when defining gRNAs with significantly altered abundance during the RRAa aggregation step, which may be specified as a single numeric value on the unit interval or as a logical vector. When supplied as a logical vector (of length equal to nrows(fit)), this parameter directly indicates the gRNAs to include during RRAa aggregation. Otherwise, if scoring is set to pvalue or combined, this parameter is interpreted as the maximum nominal p-value required to consider a gRNA's abundance meaningfully altered during the aggregation step. If scoring is fc, this parameter is interpreted as the proportion of the list to be considered meaningfully altered in the experiment (e.g., if RRAalphaCutoff is set to 0.05, only consider the rankings of the 5 (or downregulated) gRNAs for the purposes of RRAa calculations).

Note that this function uses directional tests to identify enriched or depleted targets, and when RRAalphaCutoff is provided as a logical vector the interpretation of the various aggregation statistics is going to be dependent on the specific criteria used to select reagents for inclusion.

permutations

The number of permutations to use during the RRAa aggregation step.

contrast.term

If a fit object with multiple coefficients is passed in, a string indiating the coefficient of interest.

scoring

The gRNA ranking method to use in RRAa aggregation. May take one of three values: pvalue, fc, or 'combined'. pvalue indicates that the gRNA ranking statistic should be created from the (one-sided) p-values in the fit object. fc indicates that the ranks of the gRNA coefficients should be used instead, and combined indicates that that the coefficents should be used as the ranking statistic but gRNAs are discarded in the aggregation step based on the corresponding nominal p-value in the fit object.

alt.annotation

Libraries targeting ambiguous biological elements (e.g., alternative promoters to a gene where the boundaries between elelments is contested) may contain reagents that are plausibly annotated to a finite set of possible targets. To accomodate this, users may supply an alternative reagent annotation in the form of a named list of vectors, where each list element corresponds something coercible to a to a character vector of associated targets that will ultimately be assembled into the 'geneSymbol' column of the 'resultsDF' object. Each of these character vectors should be named identically to a row of the supplied fit object (e.g., the 'row.names'). It is assumed that the 'geneID' values are assigned unambiguously to the reagents, and are passed through directly.

permutation.seed

numeric seed for permutation reproducibility. Default: NULL means to not set any seed. This argument is passed through to ct.RRAaPvals.

Value

A dataframe containing gRNA-level and target-level statistics. In addition to the information present in the supplied annotation object, the returned object indicates P-values and Q-values for the depletion and enrichment of each gRNA and associated target, the median log2 fold change estimate among all gRNAs associated with the target, and Rho statistics that are calculated internally by the RRAa algorithm that may be useful in ranking targets that are considered significant at a given alpha or false discovery threshold.

A 'resultsDF' formatted dataframe containing gene-level statistics.

Author(s)

Russell Bainer

Examples

data('fit')
data('ann')
output <- ct.generateResults(fit, ann, permutations = 10)
head(output)
  p = seq(0, 1, length.out=20)
  fc = seq(-3, 3, length.out=20)
  fc[2] = NA
  fc[3] = -20
  stats = data.frame(
    Depletion.P=p,
    Enrichment.P=rev(p),
    fc=fc
  )
  ct.applyAlpha(stats,scoring='combined')

OscarBrock/gCrisprTools documentation built on Oct. 25, 2022, 7:29 a.m.