cuffdiff_report.R

#!/usr/bin/env Rscript
#+ setup, cache=FALSE

suppressMessages(require(docopt))

doc <- '
Convert a cuffdiff into a dge-report
Usage: cuffdiff_report.R [options] <cuffdb_directory>

Options:
--constrasts=<tab_delim_table> Table with sample pairs for which dge analysis should be performed
--qcutoff <qcutoff>    Use a q-value cutoff of 0.01 instead of a q-value cutoff [default: 0.01]
--pcutoff <pcutoff>    Override q-value filter and filter by p-value instead
--minfpkm <minfpkm>   Minimal expression in any of the sample to be included in the final result list [default: 1]
'
#--directed           Perform directed dge-steps in a directed manner
#-S                     Dont do general statistics and qc analysis

#print("args ars")
#print(commandArgs(TRUE))

opts <- docopt(doc, commandArgs(TRUE)) ## does not work when spining
## florio_11b: setwd("/lustre/projects/bioinfo/holger/projects/florio_11b/cuffdiff/dge_report_new"); opts <- docopt(doc, "--constrasts contrasts.txt --pcutoff 0.01 /projects/bioinfo/holger/projects/florio_11b/cuffdiff")
# opts <- docopt(doc, "--undirected  ..")
# opts <- docopt(doc, paste(Sys.getenv("DGE_PARAMS"), ".."))
## todo use commandArgs here!!
#opts <- docopt(doc, paste(Sys.getenv("DGE_PARAMS"), paste(commandArgs(TRUE), collapse=" ")))
#print(opts)

skipQC <<- opts$S
#split_hit_list <- opts$directed
constrasts_file <- opts$constrasts
pcutoff <- if(is.null(opts$pcutoff)) NULL else as.numeric(opts$pcutoff)
qcutoff <- if(is.numeric(pcutoff)) NULL else as.numeric(opts$qcutoff)
if(is.numeric(pcutoff)) opts$qcutoff <- NULL

minFPKM <- as.numeric(opts$minfpkm)


## do basic validation of configuration
stopifnot(is.numeric(qcutoff) | is.numeric(pcutoff))

cuffdb_directory <- normalizePath(opts$cuffdb_directory)

if(is.na(file.info(cuffdb_directory)$isdir)){
    stop(paste("db directory does not exist", doc))
}


devtools::source_url("https://raw.githubusercontent.com/holgerbrandl/datautils/v1.7/R/core_commons.R")
devtools::source_url("https://raw.githubusercontent.com/holgerbrandl/datautils/v1.7/R/ggplot_commons.R")
devtools::source_url("https://raw.githubusercontent.com/holgerbrandl/datautils/v1.7/R/bio/diffex_commons.R")

#+ results='asis', echo=FALSE
cat('
<link rel="stylesheet" type="text/css" href="http://cdn.datatables.net/1.10.5/css/jquery.dataTables.min.css">
<script type="text/javascript" charset="utf8" src="http://code.jquery.com/jquery-2.1.2.min.js"></script>
<script type="text/javascript" charset="utf8" src="http://cdn.datatables.net/1.10.5/js/jquery.dataTables.min.js"></script>

<script type="text/javascript">
         $(document).ready(function() {
            // alert("test")
             $(".dtable").DataTable();
         } );
</script>
')

require(cummeRbund)
require(knitr)


geneInfo <- quote({
    martName <- guess_mart(fpkm(genes(cuff))$gene_id)
    mart <- biomaRt::useDataset(martName, mart = biomaRt::useMart("ensembl"))
    biomaRt::getBM(c('ensembl_gene_id', 'external_gene_name', 'description', 'gene_biotype'), mart=mart)
}) %>% cache_it("gene_info")

geneLoci <- quote({
    martName <- guess_mart(fpkm(genes(cuff))$gene_id)
    mart <- biomaRt::useDataset(martName, mart = biomaRt::useMart("ensembl"))
    biomaRt::getBM(c("ensembl_gene_id", "chromosome_name", "start_position", "end_position"), mart=mart)
}) %>% cache_it("gene_loci")
#knitr::opts_knit$set(root.dir = getwd())

#' Used DGE Analysis Paramters
vec2df(unlist(opts)) %>% filter(!str_detect(name, "^[<-]")) %>% kable()


########################################################################################################################
#' # Differential Gene Expression Analysis

#' [CuffDiff](http://cufflinks.cbcb.umd.edu/manual.html) and [cummeRbund](http://compbio.mit.edu/cummeRbund/) were used to perform this analysis. For details about how the test for differntial expression works, refer to [Differential analysis of gene regulation at transcript resolution with RNA-seq](http://www.nature.com/nbt/journal/v31/n1/full/nbt.2450.html).


#+ load_db, cache=FALSE, warning=FALSE
isCluster=Sys.getenv("LSF_SERVERDIR")!=""
dbTempCopyRequired=isCluster | str_detect(normalizePath(cuffdb_directory), fixed("/Volumes/projects"))

if(dbTempCopyRequired){
    ## workaround until sqlite  is fixed on lustre
    tmpdir <- tempfile("cuffdb_dir")
    system(paste("cp -r", cuffdb_directory, tmpdir))
    cuff <- readCufflinks(dir=tmpdir)
}else{
    cuff <- readCufflinks(dir="..")
}

#' The expression tracking reference contained:
cuff

#' ### Samples
runInfo(cuff) %>% filter(param!="cmd_line") %>% kable()
#' cuffdiff cmd was
#+ results='asis'
cat((runInfo(cuff) %>% filter(param=="cmd_line"))$value %>% gsub('(.{1,80})(,|$|\\s)', '\\1\n', .))

#+
replicateInfo <- replicates(cuff) %>%
    mutate(file=basename(file)) %>%
    select(-c(total_mass, norm_mass, internal_scale, external_scale))

replicateInfo %>% kable() # kable("html", table.attr = "class='dtable'", escape=F)

write.delim(replicateInfo, file="replicate_info.txt")
# replicateInfo <- read.delim("replicate_info.txt")
#'  [replicateInfo](replicate_info.txt)

#+ results='asis'
## define or load the contrasts
dbContrasts <- diffData(genes(cuff)) %>% select(sample_1, sample_2) %>% distinct()
if(!is.null(constrasts_file)){
    echo("Using contrasts matrix from: ", basename(constrasts_file))

    contrasts <- read.csv(constrasts_file, header=F) %>% set_names(c("sample_1", "sample_2"))

    ## filter for the direction present in cuffdb
    contrasts <- rbind_list(contrasts, contrasts %>% select(sample_1=sample_2, sample_2=sample_1)) %>%  inner_join(dbContrasts)
}else{
    echo("Comparing all samples against each other")
    contrasts <- diffData(genes(cuff)) %>% select(sample_1, sample_2) %>% distinct()
}

#' The used contrasts model is
contrasts %>% kable()


#######################################################################################################################
#' ## Count and Expression Score Tables

#' All data has been annotated with ensembl gene information and was exprorted into . Both normalized (FPKM) and raw counts were exported.

#+ results='asis'
fpkmMatrix(genes(cuff)) %>%
    rownames2column("ensembl_gene_id") %>%
    merge(geneInfo, by="ensembl_gene_id", all.x=T) %T>%
    head_html() %>%
    write.delim(file="gene_fpkms.txt")
# gene_fpkms <- read.delim("gene_fpkms.txt")
#' [Normalized Counts](gene_fpkms.txt)

## export the same but now including replicate information

#+ results='asis'
repFpkmMatrix(genes(cuff)) %>%
    rownames2column("ensembl_gene_id") %>%
    merge(geneInfo, by="ensembl_gene_id", all.x=T) %>%
    head_html() %>%
    write.delim(file="gene_fpkms_by_replicate.txt")
# gene_fpkms_by_replicate <- read.delim("gene_fpkms_by_replicate.txt")
#' [Normalized Counts by Replicate](gene_fpkms_by_replicate.txt)

## also export count tables
countMatrix(genes(cuff)) %>%
    rownames2column("ensembl_gene_id") %>%
    merge(geneInfo, by="ensembl_gene_id", all.x=T) %>%
    write.delim(file="gene_counts.txt")
# gene_counts <- read.delim("gene_counts.txt")
#' [Gene Counts](gene_counts.txt)

repCountMatrix(genes(cuff)) %>%
    rownames2column("ensembl_gene_id") %>%
    merge(geneInfo, by="ensembl_gene_id", all.x=T) %>%
    write.delim(file="gene_counts_by_replicate.txt")
# rep_counts_by_replicate <- read.delim("gene_counts_by_replicate.txt")
#' [Gene Counts by Replicate](gene_counts_by_replicate.txt)


#######################################################################################################################
#' ## Quality Control

#' Create some global summary statistic plots in order to get an idea about the general quaniltiy and distribution of the expression data. Descriptions are copied from [Cummerbund Manual](http://compbio.mit.edu/cummeRbund/manual_2_0.html)


#' Several plotting methods are available that allow for quality-control or global analysis of cufflinks data. A good place to begin is to evaluate the quality of the model fitting. Overdispersion is a common problem in RNA-Seq data. As of cufflinks v2.0 mean counts, variance, and dispersion are all emitted, allowing you to visualize the estimated overdispersion for each sample as a quality control measure.
dispersionPlot(genes(cuff)) #+ aes(alpha=0.01)

#' The coefficient of variation (CV) is defined as the ratio of the standard deviation the mean. The squared coefficient of variation is a normalized measure of cross-replicate variability that can be useful for evaluating the quality your RNA-seq data. Differences in CV2 can result in lower numbers of differentially expressed genes due to a higher degree of variability between replicate fpkm estimates.
fpkmSCVPlot(genes(cuff))
#fpkmSCVPlot(genes(cuff), showPool=T)

#' To assess the distributions of FPKM scores across samples, we plot their densities
csDensity(genes(cuff))
#ggsave2(csBoxplot(genes(cuff))+ ggtitle("fpkm boxplots"))

csDensity(genes(cuff),features=T)
csBoxplot(genes(cuff),replicates=T)

csScatterMatrix(genes(cuff))

#csVolcano(genes(cuff),"a", "b")
#' For the volcano plot matrix we calculate fold-changes as x/y (aka column/row). For details see [here](http://rpackages.ianhowson.com/bioc/cummeRbund/man/csVolcano.html)
csVolcanoMatrix(genes(cuff))

#' We also plot more detailed scatter plots for just the contrasts for interest. Signifcance coloring is based on cuffdiff's internal score an not our hit criterion.
#csVolcano(genes(cuff),"hESC","Fibroblasts")
csVolcPlots <- alply(contrasts, 1,  splat(function(sample_1, sample_2) {
    csVolcano(genes(cuff),sample_1,sample_2)
 }))

#+ fig.width=16
multiplot(plotlist=csVolcPlots, cols=min(3, length(csVolcPlots)))

#csScatter(genes(cuff),"hESC","Fibroblasts",smooth=T)
#csScatter(genes(cuff), "a", "d", smooth=F) + aes(color=abs(log2(a/b))>2) + scale_colour_manual(values=c("darkgreen", "red"))

#MAplot(genes(cuff), "a", "d")+ aes(color=abs(log2(M))>2)+ ggtitle("MA-plot: a/d")
#MAplot(genes(cuff), "a", "b", useCount=T) #+ aes(color=abs(log2(M))>2)+ ggtitle("MA-plot: Luc/KD")

#csVolcano(genes(cuff),"aRGm", "N") + ggtitle("odd line-shaped fan-out in volcano plot")

## todo consider to add more pretty volcano plots (see stefania's extra plots
#getGenes(cuff,c("ENSMUSG00000097853")) %>%  expressionBarplot()


#######################################################################################################################
#' ## Replicate Correlation and Clustering

#' For most type of experiments we expect replicates to cluster together. Second similar biological conditions should cluster together. As clustering results depend on the method being used, we try a few here to check if we get consistent results.

#' Map replicates into 2-dimensional space to highlight replicate clusters (or the lack of them)
require.auto(edgeR)
plotMDS(repFpkmMatrix(genes(cuff)), top=500, main="top500 MDS")


if(unlen(replicates(cuff)$sample) > 2){  ## >2 only, because MDS will fail with just 2 samples
    MDSplot(genes(cuff)) + ggtitle("mds plot")
}

## todo replace with marta plot
#PCAplot(genes(cuff),"PC1","PC2",replicates=T)

noTextLayer <- function(gg){ gg$layers[[2]] <- NULL;; gg}


## todo use fix scaled here
noTextLayer(csDistHeat(genes(cuff)) +  ggtitle("Sample correlation"))

noTextLayer(csDistHeat(genes(cuff),replicates=T) +  ggtitle("Replicate Correlation")) #+ scale_fill_gradientn(colours=c(low='lightyellow',mid='orange',high='darkred'), limits=c(0,0.15))

csDendro(genes(cuff),replicates=T)

#pdf(paste0(basename(getwd()),"_rep_clustering.pdf"))
#csDendro(genes(cuff),replicates=T)
#dev.off()

#' Also cluster only expressed genes (fpkm >1 for at least one sample)
exprGenes <- repFpkmMatrix(genes(cuff)) %>%
    rownames2column("ensembl_gene_id") %>%
    melt(value.name="fpkm") %>%
    group_by(ensembl_gene_id) %>%
    ## apply log trafo
    mutate(log_fpkm=log10(fpkm+1)) %>%
    ## apply expression filter
    filter(max(log_fpkm)>1) %>%
    dcast( ensembl_gene_id ~ variable, value.var="log_fpkm") %>%
    column2rownames("ensembl_gene_id")

res<-JSdist(makeprobs(exprGenes)) %>% hclust() %>% as.dendrogram()

par(mar=c(10,4,4,4))
plot(res, main="Replicate Clustering of Expressed Genes")


#######################################################################################################################
#' ## Differentially expressed genes

#' Here we are using [cuffdiff](http://cole-trapnell-lab.github.io/cufflinks/cuffdiff) to test for differential expression. Estimate variance-mean dependence in count data from high-throughput sequencing assays and test for differential expression based on a model using the negative binomial distribution.

allDiff <- diffData(genes(cuff)) %>%
    ## just keep pairs of interest
    merge(contrasts) %>%

    ##  discard all genes that are not expressed in any of the cell types (>1)
    filter(gene_id %in% getExpressedGenes(cuff, minFPKM=minFPKM)) %>%
    dplyr::rename(ensembl_gene_id=gene_id, s2_over_s1_log2fc=log2_fold_change) %>%
    mutate(sample_1_overex=s2_over_s1_log2fc<0)

#+ results='asis'
if(!is.null(qcutoff)){
    echo("Using q-value cutoff of", qcutoff)
    allDiff <- transform(allDiff, is_hit=q_value<=qcutoff)
}else{
    echo("Using p-value cutoff of", pcutoff)
    allDiff <- transform(allDiff, is_hit=p_value<=pcutoff)
}

merge(allDiff, geneInfo, by="ensembl_gene_id", all.x=T) %>% write.delim(file="diff_data.txt")
# diff_data <- read.delim("diff_data.txt")
#'  [Unfiltered CuffDiff Results](diff_data.txt)
#

# Filter for DEGs (differentially expressed genes)
degs <- allDiff %>%
    ## Filter for expressed genes
    filter(is_hit) %>%
    ## filter(significant=="yes") ## original hit-flag provided by cummerbund§
    ## add gene info
    left_join(geneInfo, by="ensembl_gene_id")


#'  DEGs (differentially expressed genes) are contained in the following table
write.delim(degs, file="degs.txt")
# degs <- read.delim("degs.txt")
#' [Differentially Expressed Genes](degs.txt)

#' ### DEG Counts
#with(degs, as.data.frame(table(sample_1, sample_2, sample_1_overex))) %>% filter(Freq >0) %>% kable()
degs %>% dplyr::count( sample_1, sample_2, sample_1_overex) %>% kable()


#+ fig.height=nrow(contrasts)+2
ggplot(degs, aes(paste(sample_1, "vs",  sample_2), fill=status)) + geom_bar() +xlab(NULL) + ylab("# DGEs") +ggtitle("DEGs by contrast") + coord_flip()
ggplot(degs, aes(paste(sample_1, "vs",  sample_2), fill=sample_1_overex)) + geom_bar(position="dodge") + xlab(NULL) + ylab("# DGEs") + ggtitle("DEGs by direction") + coord_flip()


#' DEGs can be browsed in Excel using the exported table or via the embedded table browser. To enable the IGV links, you need to set the port in the IGV preferences to 3334.
#+ results='asis'
degs %>%
    inner_join(geneLoci) %>%
    mutate(
        ensembl=paste0("<a href='http://www.ensembl.org/Multi/Results?page=1;facet_feature_type=Gene;q=",ensembl_gene_id, "'>",ensembl_gene_id, "</a>"),
        igv=paste0("<a href='http://localhost:3334/goto?locus=", chromosome_name,":", start_position, "-", end_position, "'>IGV</a>")
    ) %>%
    select(sample_1, sample_2, value_1, value_2, s2_over_s1_log2fc, p_value, q_value, ensembl, external_gene_name, description, gene_biotype, igv) %>%
    kable("html", table.attr = "class='dtable'", escape=F)

## just needed to restore environment easily
save(degs, file=".degs.RData")
# degs <- local(get(load(".degs.RData")))

#' Also redo volcano plots with out hit overlays

#+ fig.width=16, fig.height=14
csVolcPlots <- alply(contrasts, 1,  splat(function(sample_1, sample_2) {
    csVolcano(genes(cuff),sample_1,sample_2)
 }))

maPlots <- allDiff %>% group_by(sample_1, sample_2) %>% do(gg={ maData <-.
    ## todo why not s2_over_s1_log2fc
    maData %>% ggplot(aes(0.5*log2(value_1*value_2), log2(value_1/value_2), color=is_hit)) +
        geom_point(alpha=0.3) +
        geom_hline(yintercept=0, color="red") +
        ggtitle(with(maData[1,], paste(sample_1, "vs", sample_2)))
}) %$% gg


#+ fig.width=16
multiplot(plotlist=maPlots, cols=min(2, length(maPlots)))


##Volcano plots
hitCounts <- degs %>%
        dplyr::count(sample_1, sample_2, sample_1_overex) %>%
        dplyr::rename(hits=n) %>%
        merge(data.frame(sample_1_overex=c(T,F), x_pos=c(-3,3)))

#' For the volcano plot matrix we calculate fold-changes as x/y (aka column/row). For details see [here](http://rpackages.ianhowson.com/bioc/cummeRbund/man/csVolcano.html)
#+ fig.width=16, fig.height=14
allDiff %>% ggplot(aes(s2_over_s1_log2fc, -log10(p_value), color=is_hit)) +
#    geom_jitter(alpha=0.3, position = position_jitter(height = 0.2)) +
    geom_point(alpha=0.3) +
#    theme_bw() +
    xlim(-3.5,3.5) +
    scale_color_manual(values = c("TRUE"="red", "FALSE"="black")) +
#    ggtitle("Insm1/Ctrl") +
    ## http://stackoverflow.com/questions/19764968/remove-point-transparency-in-ggplot2-legend
    guides(colour = guide_legend(override.aes = list(alpha=1))) +

    ## tweak axis labels
    xlab(expression(log[2]("x/y"))) +
    ylab(expression(-log[10]("p value"))) +

    ## add hit couts
    geom_text(aes(label=hits, x=x_pos), y=2, color="red", size=10, data=hitCounts) +
    facet_grid(sample_1 ~ sample_2)


########################################################################################################################
#' ## Term enrichment
#+ echo=FALSE

#' This analysis was performed using [David](http://david.abcc.ncifcrf.gov/). The following ontologies were tested: `r paste(ontologies, collapse=', ')`

geneLists <-  degs %>% transmute(ensembl_gene_id, list_id=paste(sample_1, "!=", sample_2))

## also do directed analysis if we have not too many contrasts
if(nrow(contrasts)<3){
    geneLists <- rbind_list(
        geneLists,
        degs %>% filter(sample_1_overex) %>% transmute(ensembl_gene_id, list_id=paste(sample_1, ">", sample_2)),
        degs %>% filter(!sample_1_overex) %>% transmute(ensembl_gene_id, list_id=paste(sample_1, "<", sample_2))
    )
}

#+ eval=T
enrResults <- quote(geneLists %>% group_by(list_id) %>% do(davidAnnotationChart(.$ensembl_gene_id))) %>%
    cache_it(paste0("enrdata_", digest(geneLists)))

write.delim(enrResults, file="enrResults.txt")
# enrResults <- read.delim("enrResults.txt")
#'  [Enrichment Results](enrResults.txt)

#' Because David is not too strigent by default we extract just those terms for which Bonferroni<0.01
sigEnrResults <- subset(enrResults, Bonferroni <0.01)
#' results='asis'
sigEnrResults %>% kable("html", table.attr = "class='dtable'", escape=F)

write.delim(sigEnrResults, file="sigEnrResults.txt")
# sigEnrResults <- read.delim("sigEnrResults.txt")
#'  [Very Significant Terms](sigEnrResults.txt)


## plot the enrichment results
#sigEnrResults %>% group_by(Category, add=T) %>% do({
#    logPlot <- . %>% ggplot(aes(Term, PValue)) +
#	    geom_bar(stat="identity")+coord_flip() +
#	    xlab("Enriched Terms") +
#	    ggtitle(.$Category[1]) +
#	    scale_y_log10()
#	    print(logPlot)
#})

#+ include=FALSE
term_category_colors <- create_palette(unique(ac(sigEnrResults$Category)))

term_barplot_files <- sigEnrResults %>% do({
    enrResultsGrp <- .
    ## DEBUG enrResultsGrp <- sigEnrResults
    label <- enrResultsGrp$list_id[1]

#    echo("processing", label)
    logPlot <- enrResultsGrp %>%
        ## fix factor order
        mutate(Term=reorder(Term, -PValue) %>% reorder(as.integer(as.factor(Category)))) %>%
        ggplot(aes(Term, PValue, fill=Category)) +
        geom_bar(stat="identity")+
        scale_fill_manual(values = term_category_colors, drop=F, name="Ontology") +
        coord_flip() +
        xlab("Enriched Terms") +
        ggtitle(label) +
        scale_y_log10()

    fileNameLabel <- label %>%
        str_replace_all("!=", "ne") %>%
        str_replace_all(">", "gt") %>%
        str_replace_all("<", "lt") %>%
        str_replace_all(" ", "_")

#        ggsave(paste0("enrichmed_terms__", fileNameLabel, ".pdf"))
#        print(logPlot)

     tmpPng <- paste0("enrterms__", fileNameLabel, ".png")
     ggsave(tmpPng, logPlot, width=10, height = round(2+nrow(enrResultsGrp)/4), limitsize=FALSE)
     data.frame(file=tmpPng)
})

#+ results="asis"
l_ply(term_barplot_files$file, function(pngFile){ cat(paste0("<img src='", pngFile, "'><br>"))})


########################################################################################################################
#' ## Enriched KEGG Pathways

#' To understand spatio-temporal changes in gene expression better we now overlay

keggPathways <- sigEnrResults %>%
    filter(Category=="KEGG_PATHWAY") %>%
    separate(Term, c('kegg_pathway_id', 'pathway_description'), sep="\\:", remove=F) %>%
    with(kegg_pathway_id) %>% ac()

require(pathview)
require(png)


#if(nrow(keggPathways)==0){
#    echo("no enriched kegg pathways were found in the dataset")
#}else{

keggOrCode <- "mmu"

## prepare p-value data
sliceData <- allDiff %>%
    mutate(
        ## define slice labels (needed for casting)
        comparison=paste0(sample_1, "__vs__", sample_2),
        ## create normalized linear score for slice plotting
        sample_1_overex=s2_over_s1_log2fc<0,
        plot_score=-log10(q_value)*ifelse(sample_1_overex, 1, -1)
    ) %>%
    dcast(ensembl_gene_id ~ comparison, value.var="plot_score") %>%
    column2rownames("ensembl_gene_id")

#sliceData %>% head %>% kable()

data.frame(set=names(sliceData)) %>% mutate(slice_index=row_number()) %>% kable()


plot_pathway <- function(pathwayID, overlayData){
#    pathwayID="mmu04015"
#    browser()
#    echo("processing pathway", pathwayID)
    pv.out <- pathview(
            gene.data = overlayData,
            pathway.id = pathwayID,
            species = keggOrCode,
#            out.suffix = pathwayID$kegg.description,
#            out.suffix = pathwayID,
            multi.state = ncol(overlayData)>1,
#            kegg.native=F,
#            node.sum = "mean", # the method name to calculate node summary given that multiple genes or compounds are mapped to it
            limit = list(gene = c(-4,4)),
            gene.idtype="ensembl"
    )

    outfile <- paste0(pathwayID, ".pathview", ifelse(ncol(overlayData)>1, ".multi", ""), ".png")

    ## interactive plotting
#    ima <- readPNG(outfile)
#    plot.new()
#    lim <- par()
#    rasterImage(ima, lim$usr[1], lim$usr[3], lim$usr[2], lim$usr[4])

    pv.out$plotfile=outfile
    pv.out$pathway_id=pathwayID

    return(pv.out)
}

#keggPathways <- c("mmu04976", "mmu04972", "mmu04810", "mmu04520", "mmu04530", "mmu04270", "mmu04015")
#keggPathways <- c("mmu04015")



pathwayPlots <- keggPathways %>% llply(function(pathwayID){
    plot_pathway(pathwayID, sliceData)
})


save(pathwayPlots, file=".pathwayPlots.RData")
# pathwayPlots <- local(get(load("pathwayPlots.RData")))


#+ results="asis"

## simple non-clickable plots
## http://stackoverflow.com/questions/12588323/r-how-to-extract-values-for-the-same-named-elements-across-multiple-objects-of
#unlist(lapply(pathwayPlots, "[[", "plotfile"))
# unlist(lapply(pathwayPlots, "[[", "plotfile")) %>% l_ply(function(pngFile){ cat(paste0("<img src='", pngFile, "'><br>"))})

#cat("
#<style type='text/css'>
# img {
#   max-width: 100%;
# }
#</style>
#")

## todo add tooltips with additinal info
## http://stackoverflow.com/questions/5478940/how-to-create-a-html-tooltip-on-top-of-image-map

## extended version with clickable links
pathwayPlots %>% l_ply(function(plotData){
    #pngFile="mmu04015.pathview.png"
#    pathway_id=(basename(pngFile) %>% str_split_fixed("[.]", 2))[,1]
    pngFile <- plotData$plotfile
    pathway_id=plotData$pathway_id

    keggNodes <- plotData$plot.data.gene %>%
        ## remove box offset
        mutate(x=x-22, y=y-8) %>% ## todo does not work for non-gene elements
        mutate(link=paste0("http://www.kegg.jp/dbget-bin/www_bget?", keggOrCode, ":", kegg.names ))

#    "http://www.kegg.jp/dbget-bin/www_bget?mmu:16412

    ## first the image itself
#    cat(paste0("<img usemap='#", pathway_id,"' src='", pngFile, "'><br>"))
    cat(paste0("<div style='width: 2000px'><img style='float: left' usemap='#", pathway_id,"' src='", pngFile, "'></div><br>"))
    cat(paste0("<map name='", pathway_id,"'>"))
    #keggNodes %>% rowwise() %>% {curNode=.; cat(curNode$name)}

    #see http://www.html-world.de/180/image-map/
    keggNodes %>% a_ply(1, function(curNode){
        rectDef=with(curNode, paste(x, y, x+width, y+height, sep=","))
        paste0("<area href='", curNode$link,"' alt='Text' coords='",rectDef , "' shape='rect'>") %>% cat()
    })
    cat("</map>")
})

#}



#+ include=FALSE
# ########################################################################################################################
# #' ## Protein-Protein Interaction Clusters using String
#
# ## see http://www.bioconductor.org/packages/release/bioc/vignettes/STRINGdb/inst/doc/STRINGdb.pdf
#
# #' We can use the string-db web service to look for protein-protein interactions.
# #' This normally results in a large hairball and many unconnected nodes and so we can look for enriched clusters.
# #' Here we plot using (high confidence interactions) enriched clusters with a node size greater than 5 proteins for our list of candidate genes showing up in at least two replicates
#
# #+ PPI plots , fig.width=30, fig.height=30, fig.retina=2, warning=FALSE, tidy=TRUE
#
# #taxTable <- list(ENSG=9606,  ENSMUSG=10090)
# #taxId <- taxTable[[degs$gene_id[1] %>% ac() %>% str_extract("^([A-z]+)")]]
# #
# #require.auto(STRINGdb)
# #
# #string_db <- STRINGdb$new(score_threshold=400, species=taxId, input_directory="/local2/user_data/brandl/data/stringr") ## score threshold default is 400
# #example1_mapped <- string_db$map( summary_allgids_df_sighit_gt1, "supplier_gene_symbol", species= removeUnmappedRows = TRUE )
# #clustersList <- string_db$get_clusters(example1_mapped$STRING_id)
# #for(network in clustersList)
# #{
# #	if(length(network)>5)
# #	{
# #		string_db$plot_network(network, add_link=FALSE)
# #	}
# #}
#

#' Created `r format(Sys.Date(), format="%B-%d-%Y")` by `r readLines(pipe('whoami'))`