diffRegionsWithPerm: Differential Binding Estimation with Permutation
In tengmx/ComplexDiff: Differential Binding Estimation for Protein Complexes

Description Usage Arguments Details Value Examples

This function does differential binding estimation and additional permutation analysis. This function works only if multiple replicates are provided for comparison; otherwise, it is the same as diffRegions which is the main function for differential binding estimation. Permutation analysis is obtained by shuffling samples between compared conditions. Customized shuffling can also be specified by parameter permute for complicated experiment designs.

1
2
3

diffRegionsWithPerm(count, bins = NULL, meta = NULL, design, sizefac,
  rccut = 15, fccut = 0.4, gap = 2, permute = NULL, maxperm = 10,
  diffmeth = c("DESeq2", "limma", "ttest"))

`count`	A matrix of read counts or a RangedSummarizedExperiment, where columns are samples and rows are genome-wide bins. This object can be generated by function `regionReads`.
`bins`	If `count` is a read count matrix, `bins` should be provided as a GRanges object recording bins of corresponding rows in `count`. If `count` is a RangedSummarizedExperiment, this parameter will be ignored.
`meta`	If `count` is a read count matrix, this should be a DataFrame object recording sample annotations. Rows of `meta` correspond to the columns of `count`. The `design` parameter treats the column names of `meta` as variables. If `count` is a RangedSummarizedExperiment, this parameter will be ignored.
`design`	A formula object which expresses how read counts for each bin depend on the variables in `meta`, e.g. '~ group + condition' etc. Or, if `count` is a RangedSummarizedExperiment, the bins and meta objects will be extracted by `rowRanges()` and `colData()` from `count` object. By default, the last variable in `design` formula will be used to build the differential binding contrast. At most two variables are allowed if `diffmeth` is set to 'ttest'. (Details see below)
`sizefac`	A numeric vector indicating estimated size of samples for normalization purpose. This vector can be generated by function `sizeFac`.
`rccut`	A numeric cutoff on normalized count matrix using `sizefac`. If positive, only bins with normalized counts larger than `rccut` in at least one sample are selected for fold change estimate. Unlike other functions in this package, moderate cutoff would be better as too large results more false negative and too small increases the time cost. (Default: 15)
`fccut`	A numeric cutoff on smoothed log2foldchanges of bins for bump hunting of differtial binding regions. Neighbor bins with fold change larger than this value will be merged together with allowed gaps. (Default: 0.4)
`gap`	A integer specifying the gaps allowed for bin merging, in the unit of number of bins. (Default: 2)
`permute`	A matrix where each row contains shuffled indices of all samples. If `NULL`, permutation is automatically obtained by shuffling samples between compared conditions. This object can be generated by using R-package `permute` as well.
`maxperm`	Maximum number of permutations to be finished. (Default:10)
`diffmeth`	Method for statistical testing of differential binding. (Default: 'DESeq2')

Three methods are provided for significance estimation of differential binding. DESeq2 allows pseudo-estimation for comparisons without replicates; otherwise, all methods can be used for comparisons with at least two replicates. The design formula can be specified as suggested by DESeq2 and limma if these two methods are selected. For ttest, design can either contain one or two components, referring to student's t-test or paired t-test based on logarithm scaled data. For consistance with other packages, the last component in design formula is the contrast on which the final differential estimation are reported.

The default permutation works on the main contrast only (the last component in design formula). If customized permutation needed, e.g. invloving several components in design formula, users should provided the customized indecies of permutations through parameter permute.

A GRanges object containing potential regions with differential binding, as well as statistical significances as meta columns.

## load sample data
data(complex)
names(complex)

## test sample data
sizefac <- sizeFac(count=complex$counts,plot=TRUE)$sizefac
library(S4Vectors)
meta <- DataFrame(cond=c("ctr","tre"))
dr <- diffRegionsWithPerm(count=complex$counts,bins=complex$bins,
                          meta=meta,design=~cond,sizefac=sizefac)