view edgeR_Differential_Gene_Expression.xml @ 57:25b54356ecfb draft

Uploaded

<?xml version="1.0" encoding="UTF-8"?>
<tool id="edger_dge" name="edgeR: Differential Gene(Expression) Analysis">
	<description>RNA-Seq gene expression analysis using edgeR (R package)</description>
	
	<requirements>
		<requirement type="package" version="3.0.1">package_r3_withx</requirement>
		<requirement type="package" version="latest">package_biocLite_edgeR_limma</requirement>
	</requirements>
	
	<command>
		<!--
			The following script is written in the "Cheetah" language:
			http://www.cheetahtemplate.org/docs/users_guide_html_multipage/contents.html
		-->
		
		R --vanilla --slave -f $R_script '--args
			$expression_matrix
			$design_matrix
			$contrast
			
			$fdr
			
			$output_count_edgeR 
			$output_cpm
			
			/dev/null													<!-- Calculation of FPKM/RPKM should come here -->
			
			#if $output_raw_counts:
				$output_raw_counts
			#else:
				/dev/null
			#end if
			
			#if $output_MDSplot:
				$output_MDSplot
			#else:
				/dev/null
			#end if
			
			#if $output_BCVplot:
				$output_BCVplot
			#else:
				/dev/null
			#end if
			
			#if $output_MAplot:
				$output_MAplot
			#else:
				/dev/null
			#end if
			
			#if $output_PValue_distribution_plot:
				$output_PValue_distribution_plot
			#else:
				/dev/null
			#end if
			
			#if $output_hierarchical_clustering_plot:
				$output_hierarchical_clustering_plot
			#else:
				/dev/null
			#end if
			
			#if $output_heatmap_plot:
				$output_heatmap_plot
			#else:
				/dev/null
			#end if
			
			#if $output_RData_obj:
				$output_RData_obj
			#else:
				/dev/null
			#end if
			
			$output_format_images
			'
			#if $output_R:
				> $output_R 
			#else:
				> /dev/null
			#end if
			
			2> stderr.txt ; 
			
			grep -v 'Calculating library sizes from column' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			
			## Locale error messages:
			grep -v 'During startup - Warning messages' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_TIME failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_MONETARY failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_PAPER failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_MEASUREMENT failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_CTYPE failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			grep -v 'Setting LC_COLLATE failed' stderr.txt > stderr2.txt ; rm stderr.txt ; mv stderr2.txt stderr.txt ; 
			
			cat stderr.txt >&amp;2
	
	</command>
	
	<inputs>
		<param name="expression_matrix" type="data" format="tabular" label="Expression (read count) matrix" />
		<param name="design_matrix" type="data" format="tabular" label="Design matrix" hepl="Ensure your samplenames are identical to those in the expression matrix. Preferentially, create the contrast matrix using 'edgeR: Design- from Expression matrix'." />
		
		<param name="contrast" type="text" label="Contrast (biological question)" help="e.g. 'tumor-normal' or '(G1+G2)/2-G3' using the factors chosen in the design matrix. Read the 'makeContrasts' manual from Limma package for more info: http://www.bioconductor.org/packages/release/bioc/html/limma.html and http://www.bioconductor.org/packages/release/bioc/vignettes/limma/inst/doc/usersguide.pdf." />
		
		<param name="fdr" type="float" min="0" max="1" value="0.05" label="False Discovery Rate (FDR)" />
		
		<param name="outputs" type="select" label="Optional desired outputs" multiple="true" display="checkboxes">
			<option value="make_output_raw_counts">Raw counts table</option>
			<option value="make_output_MDSplot">MDS-plot</option>
			<option value="make_output_BCVplot">BCV-plot</option>
			<option value="make_output_MAplot">MA-plot</option>
			<option value="make_output_PValue_distribution_plot">P-Value distribution plot</option>
			<option value="make_output_hierarchical_clustering_plot">Hierarchical custering</option>
			<option value="make_output_heatmap_plot">Heatmap</option>
			
			<option value="make_output_R_stdout">R stdout</option>
			<option value="make_output_RData_obj">R Data object</option>
		</param>
		
		<param name="output_format_images" type="select" label="Output format of images" display="radio">
			<option value="png">Portable network graphics (.png)</option>
			<option value="pdf">Portable document format (.pdf)</option>
			<option value="svg">Scalable vector graphics (.svg)</option>
		</param>
	</inputs>
	
	<configfiles>
		<configfile name="R_script">
library(limma,quietly=TRUE) ## enable quietly to avoid unnecessaity stderr dumping
library(edgeR,quietly=TRUE) ## enable quietly to avoid unnecessaity stderr dumping
library(splines,quietly=TRUE) ## enable quietly to avoid unnecessaity stderr dumping
 
## Fetch commandline arguments
args &lt;- commandArgs(trailingOnly = TRUE)

expression_matrix_file              = args[1]
design_matrix_file                  = args[2]
contrast                            = args[3]

fdr                                 = args[4]

output_count_edgeR                  = args[5]
output_cpm                          = args[6]

output_xpkm                         = args[7]							##FPKM file - yet to be implemented

output_raw_counts                   = args[8]
output_MDSplot                      = args[9]
output_BCVplot                      = args[10]
output_MAplot                       = args[11]
output_PValue_distribution_plot     = args[12]
output_hierarchical_clustering_plot = args[13]
output_heatmap_plot                 = args[14]
output_RData_obj                    = args[15]
output_format_images                = args[16]


library(edgeR)
##raw_data &lt;- read.delim(designmatrix,header=T,stringsAsFactors=T)
## Obtain read-counts

expression_matrix &lt;- read.delim(expression_matrix_file,header=T,stringsAsFactors=F,row.names=1,check.names=FALSE,na.strings=c(""))
design_matrix &lt;- read.delim(design_matrix_file,header=T,stringsAsFactors=F,row.names=1,check.names=FALSE,na.strings=c(""))

colnames(design_matrix) &lt;- make.names(colnames(design_matrix))

for(i in 1:ncol(design_matrix)) {
  old = design_matrix[,i]
  design_matrix[,i] = make.names(design_matrix[,i])
  if(paste(design_matrix[,i],collapse="\t") != paste(old,collapse="\t")) {
    print("Renaming of factors:")
    print(old)
    print("To:")
    print(design_matrix[,i])
  }
  ## The following line seems to malfunction the script:
  ##design_matrix[,i] &lt;- as.factor(design_matrix[,i])
}

## 1) In the expression matrix, you only want to have the samples described in the design matrix
columns &lt;- match(rownames(design_matrix),colnames(expression_matrix))
columns &lt;- columns[!is.na(columns)]
read_counts &lt;- expression_matrix[,columns]

## 2) In the design matrix, you only want to have samples of which you really have the counts
columns &lt;- match(colnames(expression_matrix),rownames(design_matrix))
columns &lt;- columns[!is.na(columns)]
design_matrix &lt;- design_matrix[columns,,drop=FALSE]

## Filter for HTSeq predifined counts:
exclude_HTSeq &lt;- c("no_feature","ambiguous","too_low_aQual","not_aligned","alignment_not_unique")
exclude_DEXSeq &lt;- c("_ambiguous","_empty","_lowaqual","_notaligned")

exclude &lt;- match(c(exclude_HTSeq, exclude_DEXSeq),rownames(read_counts))
exclude &lt;- exclude[is.na(exclude)==0]
if(length(exclude) != 0)  {
  read_counts &lt;- read_counts[-exclude,]
}


empty_samples &lt;- apply(read_counts,2,function(x) sum(x) == 0)
if(sum(empty_samples) > 0) {
  write(paste("There are ",sum(empty_samples)," empty samples found:",sep=""),stderr())
  write(colnames(read_counts)[empty_samples],stderr())
} else {
  
  dge &lt;- DGEList(counts=read_counts,genes=rownames(read_counts))
  
  formula &lt;- paste(c("~0",make.names(colnames(design_matrix))),collapse = " + ")
  design_matrix_tmp &lt;- design_matrix
  colnames(design_matrix_tmp) &lt;- make.names(colnames(design_matrix_tmp))
  design &lt;- model.matrix(as.formula(formula),design_matrix_tmp)
  rm(design_matrix_tmp)
  
  # Filter prefixes
  prefixes = colnames(design_matrix)[attr(design,"assign")]
  avoid = nchar(prefixes) == nchar(colnames(design))
  replacements = substr(colnames(design),nchar(prefixes)+1,nchar(colnames(design)))
  replacements[avoid] = colnames(design)[avoid]
  colnames(design) = replacements
  
  # Do normalization
  write("Calculating normalization factors...",stdout())
  dge &lt;- calcNormFactors(dge)
  write("Estimating common dispersion...",stdout())
  dge &lt;- estimateGLMCommonDisp(dge,design)
  write("Estimating trended dispersion...",stdout())
  dge &lt;- estimateGLMTrendedDisp(dge,design)
  write("Estimating tagwise dispersion...",stdout())
  dge &lt;- estimateGLMTagwiseDisp(dge,design)
  
  
  if(output_MDSplot != "/dev/null") {
    write("Creating MDS plot",stdout())
    ##points &lt;- plotMDS(dge,method="bcv",labels=rep("",nrow(dge\$samples)))# Get coordinates of unflexible plot
    points &lt;- plotMDS.DGEList(dge,labels=rep("",nrow(dge\$samples)))# Get coordinates of unflexible plot
    dev.off()# Kill it
    
    if(output_format_images == "pdf") {
      capabilities()
      ##x11()
      pdf(output_MDSplot)
    } else if(output_format_images == "svg") {
      svg(output_MDSplot)
    } else {
      png(output_MDSplot)
    }
    
    diff_x &lt;- abs(max(points\$x)-min(points\$x))
    diff_y &lt;-(max(points\$y)-min(points\$y))
    plot(c(min(points\$x),max(points\$x) + 0.45 * diff_x), c(min(points\$y) - 0.05 * diff_y,max(points\$y) + 0.05 * diff_y), main="edgeR MDS Plot",type="n", xlab="BCV distance 1", ylab="BCV distance 2")
    points(points\$x,points\$y,pch=20)
    text(points\$x, points\$y,rownames(dge\$samples),cex=0.7,col="gray",pos=4)
    rm(diff_x,diff_y)
    
    dev.off()
  }
  
  if(output_BCVplot != "/dev/null") {
    write("Creating Biological coefficient of variation plot",stdout())
    pdf(output_BCVplot)
    plotBCV(dge, cex=0.4, main="edgeR: Biological coefficient of variation (BCV) vs abundance")
    dev.off()
  }
  
  
  write("Fitting GLM...",stdout())
  fit &lt;- glmFit(dge,design)

  write(paste("Performing likelihood ratio test: ",contrast,sep=""),stdout())
  cont &lt;- c(contrast)
  cont &lt;- makeContrasts(contrasts=cont, levels=design)

  lrt &lt;- glmLRT(fit, contrast=cont[,1])
  write(paste("Exporting to file: ",output_count_edgeR,sep=""),stdout())
  write.table(file=output_count_edgeR,topTags(lrt,n=nrow(read_counts))\$table,sep="\t",row.names=TRUE,col.names=NA)
  write.table(file=output_cpm,cpm(dge,normalized.lib.sizes=TRUE),sep="\t",row.names=TRUE,col.names=NA)

  ## todo EXPORT FPKM
  write.table(file=output_raw_counts,dge\$counts,sep="\t",row.names=TRUE,col.names=NA)
  
  
  if(output_MAplot != "/dev/null" || output_PValue_distribution_plot != "/dev/null") {
    etable &lt;- topTags(lrt, n=nrow(dge))\$table
    etable &lt;- etable[order(etable\$FDR), ]
    
    if(output_MAplot != "/dev/null") {
      write("Creating MA plot...",stdout())
      pdf(output_MAplot)
      with(etable, plot(logCPM, logFC, pch=20, main="edgeR: Fold change vs abundance"))
      with(subset(etable, FDR &lt; fdr), points(logCPM, logFC, pch=20, col="red"))
      abline(h=c(-1,1), col="blue")
      dev.off()
    }
  
    if(output_PValue_distribution_plot != "/dev/null") {
      write("Creating P-value distribution plot...",stdout())
      pdf(output_PValue_distribution_plot)
      expressed_genes &lt;- subset(etable, PValue &lt; 0.99)
      h &lt;- hist(expressed_genes\$PValue,breaks=nrow(expressed_genes)/15,main="Binned P-Values (&lt; 0.99)")
      center &lt;- sum(h\$counts) / length(h\$counts)
      lines(c(0,1),c(center,center),lty=2,col="red",lwd=2)
      k &lt;- ksmooth(h\$mid, h\$counts)
      lines(k\$x,k\$y,col="red",lwd=2)
      rmsd &lt;- (h\$counts) - center
      rmsd &lt;- rmsd^2
      rmsd &lt;- sum(rmsd)
      rmsd &lt;- sqrt(rmsd)
      text(0,max(h\$counts),paste("e=",round(rmsd,2),sep=""),pos=4,col="blue")
      ## change e into epsilon somehow
      dev.off()
    }
  }
  
  if(output_heatmap_plot != "/dev/null") {
    pdf(output_heatmap_plot,width=10.5)
    etable2 &lt;- topTags(lrt, n=100)\$table
    order &lt;- rownames(etable2)
    cpm_sub &lt;- cpm(dge,normalized.lib.sizes=TRUE,log=TRUE)[as.numeric(order),]
    heatmap(t(cpm_sub))
    dev.off()
  }
  
  ##output_hierarchical_clustering_plot = args[13]
  
  if(output_RData_obj != "/dev/null") {
    save.image(output_RData_obj)
  }
  
  write("Done!",stdout())
}
		</configfile>
	</configfiles>
	
	<outputs>
		<data format="tabular" name="output_count_edgeR" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - differentially expressed genes" />
		<data format="tabular" name="output_cpm" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - CPM" />
		
		<data format="tabular" name="output_raw_counts" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - raw counts">
			<filter>outputs and ("make_output_raw_counts" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_MDSplot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - MDS-plot">
			<filter>outputs and ("make_output_MDSplot" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_BCVplot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - BCV-plot">
			<filter>outputs and ("make_output_BCVplot" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_MAplot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - MA-plot">
			<filter>outputs and ("make_output_MAplot" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_PValue_distribution_plot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - P-Value distribution">
			<filter>outputs and ("make_output_PValue_distribution_plot" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_hierarchical_clustering_plot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - Hierarchical custering">
			<filter>outputs and ("make_output_hierarchical_clustering_plot" in outputs)</filter>
		</data>
		
		<data format="${output_format_images}" name="output_heatmap_plot" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - Heatmap">
			<filter>outputs and ("make_output_heatmap_plot" in outputs)</filter>
		</data>
		
		<data format="RData" name="output_RData_obj" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - R data object">
			<filter>outputs and ("make_output_RData_obj" in outputs)</filter>
		</data>
		
		<data format="txt" name="output_R" label="edgeR DGE on ${design_matrix.hid}: ${design_matrix.name} - R output (debug)" >
			<filter>outputs and ("make_output_R_stdout" in outputs)</filter>
		</data>
	</outputs>
	
	<help>
edgeR: Differential Gene(Expression) Analysis
#############################################

Overview
--------
Differential expression analysis of RNA-seq and digital gene expression profiles with biological replication. Uses empirical Bayes estimation and exact tests based on the negative binomial distribution. Also useful for differential signal analysis with other types of genome-scale count data [1].

For every experiment, the algorithm requires a design matrix. This matrix describes which samples belong to which groups.
More details on this are given in the edgeR manual: http://www.bioconductor.org/packages/2.12/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf
and the limma manual.

Because the creation of a design matrix can be complex and time consuming, especially if no GUI is used, this package comes with an alternative tool which can help you with it.
This tool is called *edgeR Design Matrix Creator*.
If the appropriate design matrix (with corresponding links to the files) is given,
the correct contrast ( http://en.wikipedia.org/wiki/Contrast_(statistics) ) has to be given.

If you have for example two groups, with an equal weight, you would like to compare either
"g1~g2" or "normal~cancer".

The test function makes use of a MCF7 dataset used in a study that indicates that a higher sequencing depth is not neccesairily more important than a higher amount of replaciates[2].

Input
-----
Expression matrix
^^^^^^^^^^^^^^^^^
::

  Geneid  "\t" Sample-1 "\t" Sample-2 "\t" Sample-3 "\t" Sample-4 [...] "\n"
  SMURF   "\t"      123 "\t"       21 "\t"    34545 "\t"       98  ...  "\n"
  BRCA1   "\t"      435 "\t"     6655 "\t"       45 "\t"       55  ...  "\n"
  LINK33  "\t"        4 "\t"      645 "\t"      345 "\t"        1  ...  "\n"
  SNORD78 "\t"      498 "\t"       65 "\t"       98 "\t"       27  ...  "\n"
  [...]

*Note: Make sure the number of columns in the header is identical to the number of columns in the body.*

Design matrix
^^^^^^^^^^^^^
::

  Sample    "\t" Condition "\t" Ethnicity "\t" Patient "\t" Batch "\n"
  Sample-1  "\t"     Tumor "\t"  European "\t"       1 "\t"     1 "\n"
  Sample-2  "\t"    Normal "\t"  European "\t"       1 "\t"     1 "\n"
  Sample-3  "\t"     Tumor "\t"  European "\t"       2 "\t"     1 "\n"
  Sample-4  "\t"    Normal "\t"  European "\t"       2 "\t"     1 "\n"
  Sample-5  "\t"     Tumor "\t"   African "\t"       3 "\t"     1 "\n"
  Sample-6  "\t"    Normal "\t"   African "\t"       3 "\t"     1 "\n"
  Sample-7  "\t"     Tumor "\t"   African "\t"       4 "\t"     2 "\n"
  Sample-8  "\t"    Normal "\t"   African "\t"       4 "\t"     2 "\n"
  Sample-9  "\t"     Tumor "\t"     Asian "\t"       5 "\t"     2 "\n"
  Sample-10 "\t"    Normal "\t"     Asian "\t"       5 "\t"     2 "\n"
  Sample-11 "\t"     Tumor "\t"     Asian "\t"       6 "\t"     2 "\n"
  Sample-12 "\t"    Normal "\t"     Asian "\t"       6 "\t"     2 "\n"

*Note: Avoid factor names that are (1) numerical, (2) contain mathematical symbols and preferebly only use letters.*

Contrast
^^^^^^^^
The contrast represents the biological question. There can be many questions asked, e.g.:

- Tumor-Normal
- African-European
- 0.5*(Control+Placebo) / Treated

Installation
------------

This tool requires no specific configurations. The following dependencies are installed automatically:

- R
- Bioconductor
   - limma

   - edgeR

License
-------
- R
   - GPL-2 &amp; GPL-3
- limma
    - GPL (&gt;=2)
- edgeR
     - GPL (&gt;=2)

References
----------

EdgeR
^^^^^
**[1] edgeR: a Bioconductor package for differential expression analysis of digital gene expression data.**

*Mark D. Robinson, Davis J. McCarthy and Gordon K. Smyth* - Bioinformatics (2010) 26 (1): 139-140.

- http://www.bioconductor.org/packages/2.12/bioc/html/edgeR.html
- http://dx.doi.org/10.1093/bioinformatics/btp616
- http://www.bioconductor.org/packages/release/bioc/html/edgeR.html

Test-data (MCF7)
^^^^^^^^^^^^^^^^
**[2] RNA-seq differential expression studies: more sequence or more replication?**

*Yuwen Liu, Jie Zhou and Kevin P. White* - Bioinformatics (2014) 30 (3): 301-304.

- http://www.ncbi.nlm.nih.gov/pubmed/24319002
- http://dx.doi.org/10.1093/bioinformatics/btt688

Contact
-------
The tool wrapper has been written by Youri Hoogstrate from the Erasmus Medical Center (Rotterdam, Netherlands) on behalf of the Translational Research IT (TraIT) project:
http://www.ctmm.nl/en/programmas/infrastructuren/traitprojecttranslationeleresearch

I would like to thank Hina Riaz - Naz Khan for her helpful contribution.

More tools by the Translational Research IT (TraIT) project can be found in the following repository:
http://testtoolshed.g2.bx.psu.edu/
	</help>
</tool>