comparison quality_report.xml @ 9:0d4d4f16d455 draft

"planemo upload for repository https://github.com/galaxyproteomics/tools-galaxyp/tree/master/tools/cardinal commit d008f6ea0f5c8435fb975a34cb99ea4d42c5ebd2"

8:bb9500286fe4

<tool id="cardinal_quality_report" name="MSI Qualitycontrol" version="@VERSION@.0">
    <description>
        mass spectrometry imaging QC
    </description>
    <macros>
        <import>macros.xml</import>

## in case RData input is MSImageSet:
if (class(msidata) == "MSImageSet"){
    msidata = as(msidata, "MSImagingExperiment")
    run(msidata) = "infile"
    }
print(class(msidata))
## remove duplicated coordinates
msidata <- msidata[,!duplicated(coord(msidata))]

## optional annotation from tabular file to obtain pixel groups (otherwise all pixels are considered to be one sample)

## Spectra multiplied with m/z (potential number of peaks)
numpeaks = ncol(msidata)*nrow(msidata)
## Percentage of intensities > 0
percpeaks = round(npeaks/numpeaks*100, digits=2)
## Number of empty TICs
TICs = colSums(spectra(msidata), na.rm=TRUE) 
NumemptyTIC = sum(TICs == 0)
## Median und sd TIC
medTIC = round(median(TICs), digits=1)
sdTIC = round(sd(TICs), digits=0)
## Median and sd # peaks per spectrum

            ### filter spectra for max m/z to have two vectors, which can be divided
            ### plot spatial distribution of fold change

            ## calculate mean intensity for each m/z over the ppm range; then calculate log2 foldchange
            mass1vector = colMeans(spectra(filtered_data1), na.rm =TRUE)
            mass2vector = colMeans(spectra(filtered_data2), na.rm = TRUE)
            foldchange= log2(mass1vector/mass2vector)
            fcmatrix = data.frame(coord(msidata)\$x, coord(msidata)\$y,foldchange)
            colnames(fcmatrix) = c("x", "y", "foldchange")

            print(ggplot(fcmatrix, aes(x=x, y=y, fill=foldchange))+

    abline(v=median(peakspermz), col="blue") 

    ########################## 13) Sum of intensities per m/z ##################

    ## Sum of all intensities for each m/z (like TIC, but for m/z instead of pixel)
    mzTIC = rowSums(spectra(msidata), na.rm=TRUE) ## calculate intensity sum for each m/z

    par(mfrow = c(2,1), mar=c(5,6,4,2))
    ## 13a) scatterplot
    plot_colorByDensity(mz(msidata),mzTIC,  main= "Sum of intensities per m/z", ylab ="")
    title(xlab="m/z", line=2.5)

    ########################## 14) Intensity distribution ######################

    par(mfrow = c(2,1), mar=c(5,6,4,2))

    ## 14a) Median intensity over spectra
    medianint_spectra = apply(spectra(msidata), 2, median, na.rm=TRUE) 
    plot(medianint_spectra, main="Median intensity per spectrum",las=1, xlab="Spectra index", ylab="")
    title(ylab="Median spectrum intensity", line=4)
    if (!is.null(levels(msidata\$annotation))){
        abline(v=abline_vector, lty = 3)}

            mean_matrix = cbind(mean_matrix, mean_mz_sample)}

        boxplot(log10(mean_matrix), ylab = "Log10 mean intensity per m/z", main="Log10 mean m/z intensities per annotation group", xaxt = "n")
        (axis(1, at = c(1:number_combined), labels=levels(msidata\$annotation), las=2))

        ## 14e) Heatmap of pearson correlation on mean intensities between annotation groups

        corr_matrix = mean_matrix
        corr_matrix[corr_matrix == 0] <- NA
        colnames(corr_matrix) = levels(msidata\$annotation)

        ## pearson correlation is only possible if there are at least 2 groups
        if (length(colnames)>1)
        {
            corr_matrix = cor(corr_matrix, method= "pearson",use="complete.obs")

            heatmap.parameters <- list(corr_matrix, 
            show_rownames = T, show_colnames = T,
            main = "Pearson correlation on mean intensities")
            do.call("pheatmap", heatmap.parameters)
        }
    }

    ################################## VI) Mass spectra and m/z accuracy ########################
    ############################################################################
    print("Mass spectra and m/z accuracy")

    ## replace any NA with 0, otherwise plot function will not work at all
    msidata_no_NA = msidata

    ## find three equal m/z ranges for the average mass spectra plots: 
    third_mz_range = nrow(msidata_no_NA)/3

    par(cex.axis=1, cex.lab=1, mar=c(5.1,4.1,4.1,2.1))
    print(plot(msidata_no_NA, run="infile", layout=c(2,2), strip=FALSE, main= "Average spectrum"))
    print(plot(msidata_no_NA[1:third_mz_range,], run="infile", layout=FALSE, strip=FALSE, main= "Zoomed average spectrum"))
    print(plot(msidata_no_NA[third_mz_range:(2*third_mz_range),], run="infile", layout=FALSE, strip=FALSE, main= "Zoomed average spectrum"))
    print(plot(msidata_no_NA[(2*third_mz_range):nrow(msidata_no_NA),], run="infile", layout=FALSE, strip=FALSE, main= "Zoomed average spectrum"))

    ## plot one average mass spectrum for each pixel annotation group

    if (!is.null(levels(msidata\$annotation))){
        ## print legend only for less than 10 samples

            ### find m/z with the highest mean intensity in m/z range (red line in plot 16) and calculate ppm difference for plot 17
            filtered_data = msidata_no_NA[mz(msidata_no_NA) >= inputcalibrantmasses[mass]-plusminusvalues[mass] & mz(msidata_no_NA) <= inputcalibrantmasses[mass]+plusminusvalues[mass],]

            if (nrow(filtered_data) > 0 & sum(spectra(filtered_data)) > 0){
                maxmassrow = rowMeans(spectra(filtered_data)) ## for each m/z average intensity is calculated
                maxvalue = mz(filtered_data)[which.max(maxmassrow)] ### m/z with highest average intensity in m/z range
                mzdifference = maxvalue - inputcalibrantmasses[mass] ### difference: theoretical value - closest m/z value
            ppmdifference = mzdifference/inputcalibrantmasses[mass]*1000000 ### calculate ppm for accuracy measurement
            }else{
                ppmdifference = NA

- Median intensity per spectrum: Scatter plot in which each point represents the median intensity for one spectrum. Dotted lines in the scatter plot separate spectra of different annotation groups.
- Histogram of intensities. 
- (annot) Intensities per annotation group: Same histogram as before but with colours to show the contribution of each pixel annotation group. 
- (annot) Log10 mean intensities per m/z and annotation group: For all pixels of an annotation group the log10 mean intensity for each m/z is calculated and shown as boxplot. 
- (annot) Pearson correlation between annotation groups (needs at least 2 groups) based on mean intensities and shown as heatmap.

**Mass spectra and m/z accuracy**

- Average mass spectra: First plot shows the average spectrum over the full m/z range, the other three plots zoom into the m/z axis.
- (annot) Average mass spectrum per annotation group.

9:0d4d4f16d455

<tool id="cardinal_quality_report" name="MSI Qualitycontrol" version="@VERSION@.1">
    <description>
        mass spectrometry imaging QC
    </description>
    <macros>
        <import>macros.xml</import>

## in case RData input is MSImageSet:
if (class(msidata) == "MSImageSet"){
    msidata = as(msidata, "MSImagingExperiment")
    run(msidata) = "infile"
    }

## remove duplicated coordinates
msidata <- msidata[,!duplicated(coord(msidata))]

## optional annotation from tabular file to obtain pixel groups (otherwise all pixels are considered to be one sample)

## Spectra multiplied with m/z (potential number of peaks)
numpeaks = ncol(msidata)*nrow(msidata)
## Percentage of intensities > 0
percpeaks = round(npeaks/numpeaks*100, digits=2)
## Number of empty TICs
TICs = pixelApply(msidata, sum)
NumemptyTIC = sum(TICs == 0)
## Median und sd TIC
medTIC = round(median(TICs), digits=1)
sdTIC = round(sd(TICs), digits=0)
## Median and sd # peaks per spectrum

            ### filter spectra for max m/z to have two vectors, which can be divided
            ### plot spatial distribution of fold change

            ## calculate mean intensity for each m/z over the ppm range; then calculate log2 foldchange
            mass1vector = pixelApply(filtered_data1, mean, na.rm =TRUE)
            mass2vector = pixelApply(filtered_data2, mean, na.rm = TRUE)
            foldchange= log2(mass1vector/mass2vector)
            fcmatrix = data.frame(coord(msidata)\$x, coord(msidata)\$y,foldchange)
            colnames(fcmatrix) = c("x", "y", "foldchange")

            print(ggplot(fcmatrix, aes(x=x, y=y, fill=foldchange))+

    abline(v=median(peakspermz), col="blue") 

    ########################## 13) Sum of intensities per m/z ##################

    ## Sum of all intensities for each m/z (like TIC, but for m/z instead of pixel)
    mzTIC = featureApply(msidata, sum, na.rm=TRUE) ## calculate intensity sum for each m/z

    par(mfrow = c(2,1), mar=c(5,6,4,2))
    ## 13a) scatterplot
    plot_colorByDensity(mz(msidata),mzTIC,  main= "Sum of intensities per m/z", ylab ="")
    title(xlab="m/z", line=2.5)

    ########################## 14) Intensity distribution ######################

    par(mfrow = c(2,1), mar=c(5,6,4,2))

    ## 14a) Median intensity over spectra
    medianint_spectra = pixelApply(msidata, median)
    plot(medianint_spectra, main="Median intensity per spectrum",las=1, xlab="Spectra index", ylab="")
    title(ylab="Median spectrum intensity", line=4)
    if (!is.null(levels(msidata\$annotation))){
        abline(v=abline_vector, lty = 3)}

            mean_matrix = cbind(mean_matrix, mean_mz_sample)}

        boxplot(log10(mean_matrix), ylab = "Log10 mean intensity per m/z", main="Log10 mean m/z intensities per annotation group", xaxt = "n")
        (axis(1, at = c(1:number_combined), labels=levels(msidata\$annotation), las=2))

        ## 14e) Heatmap of mean intensities of annotation groups

        colnames(mean_matrix) = levels(msidata\$annotation)
        mean_matrix[is.na(mean_matrix)] = 0
            heatmap.parameters <- list(mean_matrix, 
            show_rownames = T, show_colnames = T,
            main = "Heatmap of mean intensities per annotation group")
            par(oma=c(3,0,0,0))
            print(heatmap(mean_matrix),margins = c(10, 10))


        ## 14f) PCA of mean intensities of annotation groups

        ## define annotation by colour
        annotation_colour = rainbow(length(levels(msidata\$annotation)))[as.factor(levels(msidata\$annotation))]
        ## transform and scale dataframe
        pca = prcomp(t(mean_matrix),center=FALSE,scale.=FALSE)
        ## plot single plot
        plot(pca\$x[,c(1,2)],col=annotation_colour,pch=19)
        ## plot pca with colours for max first 5 PCs
        pc_comp = ifelse(ncol(pca\$x)<5 , ncol(pca\$x), 5)
        pairs(pca\$x[,1:pc_comp],col=annotation_colour,pch=19)
        legend("bottom", horiz = TRUE, legend=levels(msidata\$annotation), col=rainbow(length(levels(msidata\$annotation))), pch=19)

    }

    ################################## VI) Mass spectra and m/z accuracy ########################
    ############################################################################
    print("Mass spectra and m/z accuracy")

    ## replace any NA with 0, otherwise plot function will not work at all
    msidata_no_NA = msidata

    ## find three equal m/z ranges for the average mass spectra plots: 
    third_mz_range = round(nrow(msidata_no_NA)/3,0)

    par(cex.axis=1, cex.lab=1, mar=c(5.1,4.1,4.1,2.1))
    print(plot(msidata_no_NA, run="infile", layout=c(2,2), strip=FALSE, main= "Average spectrum"))
    print(plot(msidata_no_NA[1:third_mz_range,], layout=FALSE, run="infile", strip=FALSE, main="Zoomed average spectrum"))
    print(plot(msidata_no_NA[third_mz_range:(2*third_mz_range),], layout=FALSE, run="infile", strip=FALSE, main="Zoomed average spectrum"))
    print(plot(msidata_no_NA[(2*third_mz_range):nrow(msidata_no_NA),], layout=FALSE, run="infile", strip=FALSE, main="Zoomed average spectrum"))

    ## plot one average mass spectrum for each pixel annotation group

    if (!is.null(levels(msidata\$annotation))){
        ## print legend only for less than 10 samples

            ### find m/z with the highest mean intensity in m/z range (red line in plot 16) and calculate ppm difference for plot 17
            filtered_data = msidata_no_NA[mz(msidata_no_NA) >= inputcalibrantmasses[mass]-plusminusvalues[mass] & mz(msidata_no_NA) <= inputcalibrantmasses[mass]+plusminusvalues[mass],]

            if (nrow(filtered_data) > 0 & sum(spectra(filtered_data)) > 0){
                maxmassrow = featureApply(filtered_data, mean) ## for each m/z average intensity is calculated
                maxvalue = mz(filtered_data)[which.max(maxmassrow)] ### m/z with highest average intensity in m/z range
                mzdifference = maxvalue - inputcalibrantmasses[mass] ### difference: theoretical value - closest m/z value
            ppmdifference = mzdifference/inputcalibrantmasses[mass]*1000000 ### calculate ppm for accuracy measurement
            }else{
                ppmdifference = NA

- Median intensity per spectrum: Scatter plot in which each point represents the median intensity for one spectrum. Dotted lines in the scatter plot separate spectra of different annotation groups.
- Histogram of intensities. 
- (annot) Intensities per annotation group: Same histogram as before but with colours to show the contribution of each pixel annotation group. 
- (annot) Log10 mean intensities per m/z and annotation group: For all pixels of an annotation group the log10 mean intensity for each m/z is calculated and shown as boxplot. 
- (annot) Heatmap of mean intensity per m/z
- (annot) PCA of mean intensity per m/z

**Mass spectra and m/z accuracy**

- Average mass spectra: First plot shows the average spectrum over the full m/z range, the other three plots zoom into the m/z axis.
- (annot) Average mass spectrum per annotation group.