MAnorm2 (version 1.2.2)

estParamHyperChIP: The Parameter Estimation Framework of HyperChIP

Description

Given a bioCond object with which a mean-variance curve has been associated, estParamHyperChIP estimates the related parameters (i.e., the number of prior degrees of freedom and the variance ratio factor) by following the framework designed in HyperChIP.

Usage

estParamHyperChIP(
  cond,
  occupy.only = TRUE,
  prob = 0.1,
  subset = NULL,
  p_low = 0.01,
  p_up = 0.1,
  return.d0 = FALSE,
  .call = TRUE,
  ...
)

Value

By default, estParamHyperChIP returns the argument

bioCond object, whose

"df.prior" and "ratio.var" components have been updated. If return.d0 is set to TRUE, it simply returns the estimated number of prior degrees of freedom.

Arguments

cond

A bioCond object with which a mean-variance curve been associated (see also fitMeanVarCurve).

occupy.only

A logical scalar. If it is TRUE (default), only occupied genomic intervals are used for the parameter estimation process. Otherwise, all intervals are used.

prob

The proportion of the intervals with the lowest (observed) mean signal intensities that shall be used in the subsequent Winsorization procedure (see "Details").

subset

Alternatively, you can set this argument to a logical vector to directly specify the intervals to be used in the Winsorization procedure. This option overrides occupy.only and prob.

p_low, p_up

Lower- and upper-proportions of extreme values to be Winsorized (see "References"). Each of them must be strictly larger than 0, and their sum must be strictly smaller than 1.

return.d0

A logical scalar. If set to TRUE, the function simply returns the estimated number of prior degrees of freedom.

.call

Never care about this argument.

...

Further arguments to be passed to estimatePriorDfRobust.

Details

Technically, estParamHyperChIP first derives a lower quantile of the observed mean signal intensities in different genomic intervals based on the prob argument. It then selects the intervals whose mean intensities are less than or equal to the quantile. Finally, it applies the Winsorization technique to the selected intervals to finish the parameter estimation (see also "References"), by using the estimatePriorDfRobust function as the underlying engine.

estParamHyperChIP is primarily designed for coordinating with varTestBioCond to call hypervariable and lowly variable intervals across samples. See "Examples" for the workflow of a standard HyperChIP analysis.

References

Tukey, J.W., The future of data analysis. The annals of mathematical statistics, 1962. 33(1): p. 1-67.

Phipson, B., et al., Robust Hyperparameter Estimation Protects against Hypervariable Genes and Improves Power to Detect Differential Expression. Annals of Applied Statistics, 2016. 10(2): p. 946-963.

Chen, H., et al., HyperChIP: identification of hypervariable signals across ChIP-seq or ATAC-seq samples. Genome Biol, 2022. 23(1): p. 62.

See Also

bioCond for creating a bioCond object; fitMeanVarCurve for fitting a mean-variance curve and using a fit.info field to characterize it; estimatePriorDfRobust for estimating the number of prior degrees of freedom and adjusting the variance ratio factors of a set of bioConds in a robust manner; varTestBioCond for calling hypervariable and invariant intervals across ChIP-seq samples contained in a bioCond.

Examples

data(H3K27Ac, package = "MAnorm2")
attr(H3K27Ac, "metaInfo")

## Treat all the samples as independent and perform a HyperChIP analysis.
# \donttest{
# Use a pseudo-reference profile as baseline in the MA normalization
# process.
autosome <- !(H3K27Ac$chrom %in% c("chrX", "chrY"))
norm <- normalize(H3K27Ac, 4:8, 9:13, baseline = "pseudo-reference",
                  common.peak.regions = autosome)
plot(attr(norm, "MA.cor"), symbreaks = TRUE, margins = c(8, 8))

# Construct a bioCond.
cond <- bioCond(norm[4:8], norm[9:13], occupy.num = 1,
                name = "all")

# Fit a mean-variance curve by using local regression.
cond <- fitMeanVarCurve(list(cond), method = "local",
                        occupy.only = TRUE, args.lp = list(nn = 1))[[1]]
summary(cond)

# Apply the parameter estimation framework of HyperChIP.
cond <- estParamHyperChIP(cond)
summary(cond)

# Perform statistical tests and visualize the results.
res <- varTestBioCond(cond)
head(res)
hist(res$pval, breaks = 100, col = "red")
plot(res)
# }