ARA: Aggregate Region Analysis
In robinweide/GENOVA: GENOVA: expore the Hi-C

View source: R/ARMLA.R

ARA	R Documentation

Aggregate Region Analysis

Description

Extracts Hi-C matrices centered around regions and averages the results for all regions. Can take orientations of regions into account.

Usage

ARA(
  explist,
  bed,
  shift = 1e+06,
  size_bin = NULL,
  size_bp = NULL,
  outlier_filter = c(0, 1),
  anchors = NULL,
  raw = TRUE,
  strand = NULL
)

Arguments

`explist`	Either a single GENOVA `contacts` object or a list of GENOVA `contacts` objects.
`bed`	A BED-formatted `data.frame` with the following 3 columns: A `character` giving the chromosome names. An `integer` with start positions. An `integer` with end positions. Note that rows wherein the second column is larger than the third column are considered to be in the reverse direction.
`shift`	An `integer` of length 1 indicating how many basepairs the anchors should be shifted. Essentially performs circular permutation of `size` for a reasonable estimate of background. The argument is ignored when `shift <= 0`.
`size_bin`, `size_bp`	The size of the lookup regions in bins (i.e. a score of 21 yields an output with 10 Hi-C bins both up- and downstream of the anchor). When `NULL` (default), it is internally set to `21` when the `size_bp` is also `NULL`. `size_bp` is an alternative parametrisation for the lookup regions, expressed in basepairs. `size_bp` is not used when the argument `size_bin` is set.
`outlier_filter`	A `numeric` of length 2 between `[0-1]` indicating quantiles of data to be used as thresholds. Data outside these thresholds are set to the nearest threshold. Setting this to `c(0, 1)` performs no outlier correction.
`anchors`	(Optional) A `matrix` with two columns containing pre-computed anchor indices. If this is set, skips calculation of anchor indices and uses this argument instead. See `anchors_ARA()`.
`raw`	A `logical` of length 1: should the raw array underlying the summary matrices be returned in the output? Should be `TRUE` if the intention is to use the `quantify` function.
`strand`	A `character` of the length `nrow(bed)`. Overrules an attempt to infer strand from `start > end` information.

Details

By default, ARA also calculates the results for shifted anchors and normalises the "obsexp" slot by off-diagonal bands.

The 'bed' argument can take in oriented entries, wherein entries with a start site larger than the end site are considered to be in the reverse direction. The reverse sites are flipped during analysis, so the orientation is the same as in the forward sites.

Value

An ARA_discovery object containing the following slots:

obsexp: An array with the dimensions size_bin x size_bin x length(explist) containing fold change values for the signal over the (banded) median, shifted values.
signal: An array with the dimensions size_bin x size_bin x length(explist) containing mean contact values for bins surrounding the anchors.
signal_raw: A list with length(explist) elements for each contacts object, wherein an element is an n x size_bin x size_bin array with contact values for each anchor. 'n' is the number of non-empty valid anchors.
shifted: An array with the dimensions size_bin x size_bin x length(explist) containing mean contact values for bins that are shift basepairs away from the anchors.
shifted_raw: A list with length(explist) elements for each contacts object, wherein an alemeent is an n x size_bin x size_bin array with contact values for each shifted anchors. 'n' is the number of non-empty valid (unshifted) anchors.

Resolution recommendation

10kb-40kb

Examples

## Not run: 
# Typical usage
ara <- ARA(list(WT_20kb, KO_20kb), ctcf_sites)

# Alternative usage with pre-calculated anchors
anchors <- anchors_ARA(WT_20kb$IDX, ctcf_sites)
ara <- ARA(list(WT_20kb, KO_20kb))

# Visualisation
visualise(ara)

## End(Not run)

robinweide/GENOVA documentation built on March 14, 2024, 11:16 p.m.