The NOISeq tutorial can be downloaded from Bioconductor website. Click here.

In NOISeq package, you can find not only the method to compute differential expression between two experimental conditions from RNA-seq data, but also other functions in order to learn more about saturation, contamination or other biases in your data as well as to explore the differential expression results.

Please, follow these instructions to run NOISeqBIO:

1) Install NOISeq Bioconductor package (see Downloads) and load it: library(NOISeq).

2) Download the zip file containing the R scripts for NOISeqBIO (from Downloads) and unzip it.

3) To load the NOISeqBIO functions in R from your working directory:

# Remember you have to change this path if you are not in the same directory:
> mypath = "NOISeqBIO_v00"  

> source(file.path(mypath, "auxiliar.R"))
> source(file.path(mypath, "MDbio.R"))
> source(file.path(mypath, "allMDbio.R")) 
> source(file.path(mypath, "noiseqbio.R"))
> source(file.path(mypath, "classes.R"))
> source(file.path(mypath, "normalization.R"))
> source(file.path(mypath, "fewreplicates.R"))

4) Read your data as in NOISeq (see the following example with Marioni's data, where liver and kidney tissues are compared) to convert them to a NOISeq object:

> data(Marioni)
> mydata = readData(data = mycounts, factors = myfactors)

5) To apply NOISeqBIO with TMM normalization (you can choose the same normalization procedures as in NOISeq):

> mynoiseq.test = noiseqbio(mydata, norm = "tmm", factor = "Tissue", r = 10)     

The parameter r is the number of permutations of sample labels to generate null distribution.

6) Finally, select the differentially expressed genes:

# Probability cutoff. It is equivalent to a FRD (adjusted p-value) of 0.05
> myq = 0.95  
> mydeg = mynoiseq.test@results[[1]][which(mynoiseq.test@results[[1]][,"prob"] > myq),]
> nrow(mydeg)
[1] 2837
> head(mydeg)
                     Kidney      Liver      theta      prob
ENSG00000187642  12.8606633   4.332902  0.8107475 0.9601231
ENSG00000188290  17.7544458   4.961376  1.3160369 0.9999999
ENSG00000187608  19.2702201  34.329320 -0.7153955 0.9543329
ENSG00000188157 691.1874857 141.575959  5.9237953 1.0000000
ENSG00000186891   0.8230924   2.736435 -0.7348520 0.9568449
ENSG00000078808 372.5016774 483.708484 -1.0066939 0.9787270

Please, remember that this is a toy example and only about 5000 genes are considered, so the results may not be trustable.

Unfortunately, the DE plots in NOISeq package cannot be used yet on these DE results. Please, find below how to draw a plot to illustrate the DE results:

> cond1 = log2(rowMeans(mycounts[,grep("Kidney", colnames(mycounts))]) + 1)
> cond2 = log2(rowMeans(mycounts[,grep("Liver", colnames(mycounts))]) + 1)
> plot(cond1, cond2, cex = 0.7, xlab = "Kidney (Average expression in log-scale)", 
       ylab = "Liver (Average expression in log-scale)", main = "NOISeqBIO on Marioni's data")
> points(cond1[rownames(mydeg)], cond2[rownames(mydeg)], pch = 20, col = "red", cex = 0.5)