noiseqbio: Differential expression method for biological replicates

Description

noiseqbio computes differential expression between two experimental conditions from read count data (e.g. RNA-seq).

Usage

noiseqbio(input, k = 0.5, norm = c("rpkm","uqua","tmm","n"), nclust = 15, plot = FALSE, factor=NULL, conditions = NULL, lc = 0, r = 50, adj = 1.5, a0per = 0.9, random.seed = 12345, filter = 1, depth = NULL, cv.cutoff = 500, cpm = 1)

Arguments

input

Object of eSet class coming from readData function or other R packages such as DESeq.

Counts equal to 0 are replaced by k. By default, k = 0.5.

norm

Normalization method. It can be one of "rpkm" (default), "uqua" (upper quartile), "tmm" (trimmed mean of M) or "n" (no normalization).

factor

A string indicating the name of factor whose levels are the conditions to be compared.

conditions

A vector containing the two conditions to be compared by the differential expression algorithm (needed when the factor contains more than 2 different conditions).

Length correction is done by dividing expression by length^lc. By default, lc = 0.

Number of permutations to generate noise distribution by resampling.

adj

Smoothing parameter for the Kernel Density Estimation of noise distribution. Higher values produce smoother curves.

nclust

Number of clusters for the K-means algorithm. Used when the number of replicates per condition is less than 5.

plot

If TRUE, a plot is generated showing the mixture distribution (f) and the noise distribution (f0) of theta values.

a0per

M and D values are corrected for the biological variability by being divided by S + a0, where S is the standard error of the corresponding statistic and a0 is determined by the value of a0per parameter. If a0per is NULL, a0 = 0. If a0per is a value between 0 and 1, a0 is the a0per percentile of S values for all features. If a0per = "B", a0 takes the highest value given by 100*max(S).

random.seed

Random seed. In order to get the same results in different runs of the method (otherwise the resampling procedure would produce different resulst), the random seed is set to this parameter value.

filter

Method to filter out low count features before computing differential expression analysis. If filter=0, no filtering is performed. If 1, CPM method is applied. If 2, Wilcoxon test method (not recommended when the number of replicates per condition is less than 5), If 3, proportion test method. Type ?filtered.data for more details.

depth

Sequencing depth of each sample to be used by filtering method. It must be data provided when the data is already normalized and filtering method 3 is to be applied.

cv.cutoff

Cutoff for the coefficient of variation per condition to be used in filtering method 1.

cpm

Cutoff for the counts per million value to be used in filtering methods 1 and 3.

Value

The function returns an object of class Output

References

Bullard J.H., Purdom E., Hansen K.D. and Dudoit S. (2010) Evaluation of statistical methods for normalization and differential expression in mRNA-seq experiments. BMC Bioinformatics 11(1):94+.

Mortazavi A., Williams B.A., McCue K., Schaeer L. and Wold B. (2008) Mapping and quantifying mammalian transcriptomes by RNA-seq. Nature Methods 5(7):621-628.

Robinson M.D. and Oshlack A. (2010) A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biology 11(3):R25+.

Marioni, J.C. and Mason, C.E. and Mane, S.M. and Stephens, M. and Gilad, Y. (2008) RNA-seq: an assessment of technical reproducibility and comparison with gene expression arrays. Genome Research, 18: 1509--1517.

Examples

Run this code

## Load the input object from Marioni's data as returned by readData()
data(myCounts)

## Computing differential expression probability by NOISeqBIO using factor "Tissue" (data will be RPKM-normalized)
mynoiseqbio = noiseqbio(mydata, k = 0.5, norm = "rpkm", factor="Tissue", lc = 1, r = 50, adj = 1.5, plot = FALSE,
                        a0per = 0.9, random.seed = 12345, filter = 1, cv.cutoff = 500, cpm = 1)