MAnorm2 (version 1.2.2)

setWeight: Set the Weights of Signal Intensities Contained in a bioCond

Description

setWeight modifies the relative precisions of signal intensities stored in a bioCond object. One typically uses this function in the form of x <- setWeight(x, weight), where x is a bioCond object and weight is a matrix of positive weights.

Usage

setWeight(x, weight = NULL, strMatrix = NULL)

Value

A bioCond object with an updated strMatrix

field. To be noted, information about the mean-variance dependence of the original bioCond object, if any, will be removed in the returned bioCond. You can re-fit it by, for example, calling

fitMeanVarCurve.

Arguments

x

A bioCond object.

weight

A matrix or data frame specifying the relative precisions of signal intensities contained in x. Must have the same number of columns as x$norm.signal. A vector is interpreted as a matrix having a single row. Note that rows of weight are recycled if necessary. By default, the same weight is assigned to each measurement in x$norm.signal.

strMatrix

An optional list of symmetric matrices specifying directly the structure matrix of each genomic interval. Elements of it are recycled if necessary. This argument, if set, overrides the weight argument. See "Details" for more information about structure matrix.

Warning

Do not directly modify the strMatrix field in a bioCond object. Instead, use this function.

Details

For each genomic interval in a bioCond object, MAnorm2 models the signal intensities of it as having a common mean and a covariance matrix proportional to the interval's structure matrix. Put it formally, \(cov(Xi | ti) = ti * Si\), where \(Xi\) is the vector of signal intensities of the \(i\)th interval, \(ti\) is a positive scalar quantifying the variation level of this interval and \(Si\) is a symmetric matrix denoting the interval's structure matrix.

Naturally, assuming there are no correlations between ChIP-seq samples, each \(Si\) is a diagonal matrix, with its diagonal elements being the reciprocal of the corresponding weights.

The structure matrices will be used to derive the sample mean and sample variance (i.e., estimate of \(ti\)) of signal intensities of each interval, using the GLS (generalized least squares) estimation. See also fitMeanVarCurve for modeling their relationship across intervals.

References

Tu, S., et al., MAnorm2 for quantitatively comparing groups of ChIP-seq samples. Genome Res, 2021. 31(1): p. 131-145.

See Also

bioCond for creating a bioCond object based on normalized signal intensities; fitMeanVarCurve for fitting the mean-variance trend across genomic intervals.

Examples

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

## Set the weights of replicate ChIP-seq samples in a bioCond.

# Construct a bioCond object for the GM12891 cell line. By default, all the
# ChIP-seq samples belonging to the bioCond have the same weight for
# estimating the mean signal intensities of genomic intervals in the cell
# line.
norm <- normalize(H3K27Ac, 5:6, 10:11)
GM12891 <- bioCond(norm[5:6], norm[10:11], name = "GM12891")

# Now we set the weight of the 2nd sample to half of the 1st one.
GM12891_2 <- setWeight(GM12891, weight = c(1, 0.5))

# Equivalently, you can achieve the same effect by setting the strMatrix
# parameter.
GM12891_3 <- setWeight(GM12891, strMatrix = list(diag(c(1, 2))))