calculateBsFoldChange: Compute fold-changes per binding site

In ZarnackGroup/BindingSiteFinder: Binding site defintion based on iCLIP data

Compute fold-changes per binding site

Description

Given count data for binding sites and background regions this function will compute fold-changes between two condition for each binding site. Computation is based on DESeq using the Likelihood ratio test to disentangle transcription level changes from binding site level changes.

Usage

calculateBsFoldChange(
  object,
  fitType = "local",
  sfType = "ratio",
  minReplicatesForReplace = 10,
  independentFiltering = TRUE,
  alpha = 0.05,
  pAdjustMethod = "BH",
  minmu = 0.5,
  filterFun = NULL,
  use.lfc.shrinkage = FALSE,
  type = c("ashr", "apeglm", "normal"),
  svalue = FALSE,
  apeAdapt = TRUE,
  apeMethod = "nbinomCR",
  match.geneID = "geneID",
  quiet = TRUE,
  veryQuiet = FALSE,
  replaceNegative = FALSE,
  removeNA = FALSE
)

Arguments

object	a BSFDataSet object
fitType	either "parametric", "local", "mean", or "glmGamPoi" for the type of fitting of dispersions to the mean intensity. See DESeq for more details.
sfType	either "ratio", "poscounts", or "iterate" for the type of size factor estimation. See DESeq for more details.
minReplicatesForReplace	the minimum number of replicates required in order to use replaceOutliers on a sample. See DESeq for more details.
independentFiltering	logical, whether independent filtering should be applied automatically. See results for more details.
alpha	the significance cutoff used for optimizing the independent filtering. See results for more details.
pAdjustMethod	he method to use for adjusting p-values. See results for more details.
minmu	lower bound on the estimated count. See results for more details.
filterFun	an optional custom function for performing independent filtering and p-value adjustment. See results for more details.
use.lfc.shrinkage	logical; whether to compute shrunken log2 fold changes for the DESeq results. See lfcShrink for more details.
type	if 'ashr', 'apeglml' or 'normal' should be used for fold change shrinkage. See lfcShrink for more details.
svalue	logical, should p-values and adjusted p-values be replaced with s-values when using apeglm or ashr. See lfcShrink for more details.
apeAdapt	logical, should apeglm use the MLE estimates of LFC to adapt the prior. See lfcShrink for more details.
apeMethod	what method to run apeglm, which can differ in terms of speed. See lfcShrink for more details.
match.geneID	character; the name of the column with the gene ID in the binding sites meta columns used for matching binding sites to genes
quiet	logical; whether to print messages or not
veryQuiet	logical; whether to print messages or not
replaceNegative	logical; force negative counts to be replaces by 0. Be careful when using this, having negative counts can point towards problems with the gene annotation in use.
removeNA	logical; force binding sites with any NA value to be removed. Be careful when using this, having negative counts can point towards problems with the gene annotation in use.

Details

Fold-changes per binding sites are corrected for transcript level changes. Essentially, background counts are used to model transcript level changes, which are then used to compute fold-changes per binding site, which are corrected for the observed transcript level changes. This is done by using a Likelihood ratio test comparing the full model (~condition + type + condition:type) to the reduced model (~condition + type).

Fold-changes for the transcript level changes are modeled explicitly in a second round of the DESeq workflow, using the default Wald test to compare changes between the conditions (~condition). Counts attributed to binding sites are removed from the gene level counts.

Results from both calculation rounds can be filtered and further manipulated with parameters given in the DESeq2 framework (see results, lfcShrink).

This function is intended to be used right after a call of filterBsBackground.

Value

a BSFDataSet object, with results from the DESeq analysis added to the meta columns of the binding site ranges.

See Also

calculateBsBackground, filterBsBackground, plotBsBackgroundFilter, DESeq

Examples

# load clip data
files <- system.file("extdata", package="BindingSiteFinder")
load(list.files(files, pattern = ".rda$", full.names = TRUE))
load(list.files(files, pattern = ".rds$", full.names = TRUE)[1])

# make example dataset
bds = makeBindingSites(bds, bsSize = 7)
bds = assignToGenes(bds, anno.genes = gns)
bds = imputeBsDifferencesForTestdata(bds)
bds = calculateBsBackground(bds, anno.genes = gns)

# calculate fold changes - no shrinkage
bds = calculateBsFoldChange(bds)

# calculate fold changes - with shrinkage
bds = calculateBsFoldChange(bds, use.lfc.shrinkage = TRUE)

ZarnackGroup/BindingSiteFinder documentation built on May 2, 2024, 12:38 a.m.