,
Stefano Pagnotta [aut, cph] | Title: | Asymmetric Within-Sample Transformation |
|---|---|
| Description: | We propose an Asymmetric Within-Sample Transformation (AWST) to regularize RNA-seq read counts and reduce the effect of noise on the classification of samples. AWST comprises two main steps: standardization and smoothing. These steps transform gene expression data to reduce the noise of the lowly expressed features, which suffer from background effects and low signal-to-noise ratio, and the influence of the highly expressed features, which may be the result of amplification bias and other experimental artifacts. |
| Authors: | Davide Risso [aut, cre, cph] ,
Stefano Pagnotta [aut, cph] |
| Maintainer: | Davide Risso <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 1.13.0 |
| Built: | 2024-07-02 04:49:32 UTC |
| Source: | https://github.com/bioc/awst |
This function implements the asymmetric within-sample transformation described in Risso and Pagnotta (2019). The function includes two steps: a standardization step and a asymmetric winsorization step. See details.
## S4 method for signature 'matrix' awst(x, poscount = FALSE, full_quantile = FALSE, sigma0 = 0.075, lambda = 13) ## S4 method for signature 'SummarizedExperiment' awst( x, poscount = FALSE, full_quantile = FALSE, sigma0 = 0.075, lambda = 13, expr_values = "counts", name = "awst" )## S4 method for signature 'matrix' awst(x, poscount = FALSE, full_quantile = FALSE, sigma0 = 0.075, lambda = 13) ## S4 method for signature 'SummarizedExperiment' awst( x, poscount = FALSE, full_quantile = FALSE, sigma0 = 0.075, lambda = 13, expr_values = "counts", name = "awst" )
x |
a matrix of (possibly normalized) RNA-seq read counts or a 'SummarizedExperiment'. |
poscount |
a logical value indicating whether positive counts only should be used for the standardization step. |
full_quantile |
a logical value indicating whether the data have been normalized with the full-quantile normalization. In this case, computations can be sped up. |
sigma0 |
a multiplicative constant to be applied to the smoothing function. |
lambda |
a parameter that controls the growth rate of the smoothing function. |
expr_values |
integer scalar or string indicating the assay that contains the matrix to use as input. |
name |
string specifying the name of the assay to be used to store the results of the transformation. |
The standardization step is based on a log-normal distribution of the high-intensity genes. Optionally, only positive counts can be used in this step (this option is especially useful for single-cell data). The winsorization step is controlled by two parameters, sigma0 and lambda, which control the growth rate of the winsorization function.
if 'x' is a matrix, it returns a matrix of transformed values, with genes in rows and samples in column. If 'x' is a 'SummarizedExperiment', it returns a 'SummarizedExperiment' with the transformed value in the 'name' slot.
matrix: the input is a matrix of (possibly normalized) counts
SummarizedExperiment: the input is a SummarizedExperiment with (possibly
normalized) counts in one of its assays.
Risso and Pagnotta (2019). Within-sample standardization and asymmetric winsorization lead to accurate classification of RNA-seq expression profiles. Manuscript in preparation.
x <- matrix(data = rpois(100, lambda=5), ncol=10, nrow=10) awst(x)x <- matrix(data = rpois(100, lambda=5), ncol=10, nrow=10) awst(x)
This function filters out genes that show a low heterogeneity, as measured by Shannon's entropy.
## S4 method for signature 'matrix' gene_filter( x, from = min(x, na.rm = TRUE), to = max(x, na.rm = TRUE), nBins = 20, heterogeneity_threshold = 0.1 ) ## S4 method for signature 'SummarizedExperiment' gene_filter( x, from = min(assay(x, awst_values), na.rm = TRUE), to = max(assay(x, awst_values), na.rm = TRUE), nBins = 20, heterogeneity_threshold = 0.1, awst_values = "awst" )## S4 method for signature 'matrix' gene_filter( x, from = min(x, na.rm = TRUE), to = max(x, na.rm = TRUE), nBins = 20, heterogeneity_threshold = 0.1 ) ## S4 method for signature 'SummarizedExperiment' gene_filter( x, from = min(assay(x, awst_values), na.rm = TRUE), to = max(assay(x, awst_values), na.rm = TRUE), nBins = 20, heterogeneity_threshold = 0.1, awst_values = "awst" )
x |
a matrix of transformed gene expression counts (typically the
results of |
from |
the minimum value from which to start binning data. |
to |
the maximum value for the binning of the data. |
nBins |
the number of bins. |
heterogeneity_threshold |
the trheshold used for the filtering. |
awst_values |
integer scalar or string indicating the assay that contains the awst-transformed values to use as input. |
Shannon's entropy is computed on the categorized data after AWST transformation. Those genes that show a lower entropy than the predefined threshold are deemed to carry too low information to be useful for the classification of the samples, and are hence removed.
if 'x' is a matrix, it returns a filtered matrix. If 'x' is a 'SummarizedExperiment', it returns a filtered 'SummarizedExperiment'
matrix: the input is a matrix of awst-transformed values.
SummarizedExperiment: the input is a SummarizedExperiment with
awst-transformed values in one of its assays.
Risso and Pagnotta (2019). Within-sample standardization and asymmetric winsorization lead to accurate classification of RNA-seq expression profiles. Manuscript in preparation.
set.seed(222) x <- matrix(rpois(75, lambda=5), ncol=5, nrow=15) a <- awst(x) gene_filter(a)set.seed(222) x <- matrix(rpois(75, lambda=5), ncol=5, nrow=15) a <- awst(x) gene_filter(a)