annotate classification.xml @ 19:4c177985028a draft default tip

planemo upload for repository https://github.com/galaxyproteomics/tools-galaxyp/tree/master/tools/cardinal commit 91e77c139cb3b7c6d67727dc39140dd79355fa0c

<tool id="cardinal_classification" name="MSI classification" version="@TOOL_VERSION@+galaxy@VERSION_SUFFIX@" profile="22.05">
    <description>spatial classification of mass spectrometry imaging data</description>
    <macros>
        <import>macros.xml</import>
    </macros>
    <expand macro="requirements"/>
    <command detect_errors="exit_code">
        <![CDATA[

        @INPUT_LINKING@
        cat '${MSI_segmentation}' &&
        Rscript '${MSI_segmentation}'

    ]]>
    </command>
    <configfiles>
        <configfile name="MSI_segmentation"><![CDATA[


################################# load libraries and read file #########################

library(Cardinal)
library(gridExtra)
library(ggplot2)
library(scales)



@READING_MSIDATA@


msidata = as(msidata, "MSImagingExperiment")

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

@DATA_PROPERTIES_INRAM@


######################################## PDF ###################################
################################################################################
################################################################################

Title = "Prediction"

#if str( $type_cond.type_method) == "training":

        Title = "$type_cond.method_cond.class_method"
#end if

pdf("classificationpdf.pdf", fonts = "Times", pointsize = 12)
plot(0,type='n',axes=FALSE,ann=FALSE)


title(main=paste0(Title," for file: \n\n", "$infile.display_name"))


##################### I) numbers and control plots #############################
################################################################################

## table with values
grid.table(property_df, rows= NULL)

int_matrix = as.matrix(spectra(msidata))
NAcount = sum(is.na(int_matrix))


if (npeaks > 0 && NAcount==0){

    opar <- par()

    ######################## II) Training #######################################
    #############################################################################
    #if str( $type_cond.type_method) == "training":
        print("training")


        ## load y response (will be needed in every training scenario)

        y_tabular = read.delim("$type_cond.annotation_file", header = $type_cond.tabular_header, stringsAsFactors = FALSE)

        #if str($type_cond.column_fold) == "None":
        y_input = y_tabular[,c($type_cond.column_x, $type_cond.column_y, $type_cond.column_response)]
        #else
        y_input = y_tabular[,c($type_cond.column_x, $type_cond.column_y, $type_cond.column_response, $type_cond.column_fold)]
        #end if
        colnames(y_input)[1:2] = c("x", "y")

        ## merge with coordinate information of msidata
        msidata_coordinates = cbind(coord(msidata)[,1:2], c(1:ncol(msidata)))
        colnames(msidata_coordinates)[3] = "pixel_index"
        merged_response = as.data.frame(merge(msidata_coordinates, y_input, by=c("x", "y"), all.x=TRUE))
        merged_response[is.na(merged_response)] = "NA"
        merged_response = merged_response[order(merged_response\$pixel_index),]
        conditions = as.factor(merged_response[,4])
        y_vector = conditions

        ## colours selection:

	    #if str($colour_conditional.colour_type) == "manual_colour"
	        #set $color_string = ','.join(['"%s"' % $color.annotation_color for $color in $colour_conditional.colours])
	        colourvector = c($color_string)

	    #elif str($colour_conditional.colour_type) == "colourpalette"
	        number_levels = (length(levels(conditions)))
	        colourvector = noquote($colour_conditional.palettes)(number_levels)

	    #end if


        ## plot of y vector

        position_df = as.data.frame(cbind(coord(msidata)[,1:2], conditions))
        y_plot = ggplot(position_df, aes(x=x, y=y, fill=conditions))+
           geom_tile() +
           coord_fixed()+
           ggtitle("Distribution of the conditions")+
		   theme_bw()+
           theme(
	       plot.background = element_blank(),
	       panel.grid.major = element_blank(),
	       panel.grid.minor = element_blank())+
           theme(text=element_text(family="ArialMT", face="bold", size=15))+
           theme(legend.position="bottom",legend.direction="vertical")+
           guides(fill=guide_legend(ncol=4,byrow=TRUE))+
           scale_discrete_manual(aesthetics = c("colour", "fill"), values = colourvector)
        coord_labels = aggregate(cbind(x,y)~conditions, data=position_df, mean, na.rm=TRUE, na.action="na.pass")
        ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$conditions)
        print(y_plot)

        ## plot of folds

        #if str($type_cond.column_fold) != "None":
            fold_vector = as.factor(merged_response[,5])

            position_df = as.data.frame(cbind(coord(msidata)[,1:2], fold_vector))
            fold_plot = ggplot(position_df, aes(x=x, y=y, fill=fold_vector))+
               geom_tile() +
               coord_fixed()+
               ggtitle("Distribution of the fold variable")+
	           theme_bw()+
               theme(
	           plot.background = element_blank(),
	           panel.grid.major = element_blank(),
	           panel.grid.minor = element_blank())+
               theme(text=element_text(family="ArialMT", face="bold", size=15))+
               theme(legend.position="bottom",legend.direction="vertical")+
               guides(fill=guide_legend(ncol=4,byrow=TRUE))
            coord_labels = aggregate(cbind(x,y)~fold_vector, data=position_df, mean, na.rm=TRUE, na.action="na.pass")
            ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$fold_vector)
            print(fold_plot)

        #end if

        ######################## PLS #############################
        #if str( $type_cond.method_cond.class_method) == "PLS":
            print("PLS")

            ######################## PLS - CV #############################
            #if str( $type_cond.method_cond.analysis_cond.PLS_method) == "cvapply":
                print("PLS cv")

                ## set variables for components and number of response groups
                components = c($type_cond.method_cond.analysis_cond.plscv_comp)
                number_groups = length(levels(y_vector))

                ## PLS-cvApply:
                msidata.cv.pls <- crossValidate(msidata, .y = y_vector,  .fold = fold_vector, .fun = "PLS", ncomp = components)

                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()

                ## create new summary table with cv results
                results_list <- NULL
                for (i in seq_along(components)) {
                  ## extract accuracy, sensitivity, and specificity for the current i
                  accuracy <- round(as.data.frame(msidata.cv.pls@resultData@listData[[i]][["accuracy"]]), digits=2)
                  sensitivity <- round(as.data.frame(msidata.cv.pls@resultData@listData[[i]][["sensitivity"]]), digits=2)
                  specificity <- round(as.data.frame(msidata.cv.pls@resultData@listData[[i]][["specificity"]]), digits=2)

                  ## combine accuracy, sensitivity, and specificity into one data frame
                  result_df <- cbind(folds = rownames(accuracy), ncomp = i, accuracy, sensitivity, specificity)
                  colnames(result_df) <- c("folds", "ncomp", "accuracy", "sensitivity", "specificity")
                  rownames(result_df) <- NULL

                  ## add column names with ncomp as first row to each dataframe
                  col_names_row <- data.frame(folds = "folds", ncomp = paste0("ncomp", i), accuracy = "accuracy", sensitivity = "sensitivity", specificity = "specificity")
                  result_df <- rbind(col_names_row, result_df)

                  results_list[[i]] <- result_df
                }

                ## combine all data frames in the list into one data frame
                results_df <- do.call(rbind, results_list)

                summary_df <- results_df

                ## new table and plot of accuracies over all components
                summary.cv.pls = as.data.frame(summary(msidata.cv.pls))
                plot(0,type='n',axes=FALSE,ann=FALSE)
                summary.cv.pls.round <- round(summary.cv.pls, digits=2)
                grid.table(summary.cv.pls.round, rows=NULL)

                accuracy_plot = ggplot(summary.cv.pls, aes(x = ncomp, y = Accuracy)) +
                  geom_point(color = "blue", size = 3) +   # Add points
                  geom_line() +
                  theme_bw()
                print(accuracy_plot)

                ## print table with summary in pdf
                par(opar)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                title(main="Summary for the different components\n", adj=0.5)
                ## 20 rows fits in one page:
                if (nrow(summary_df)<=20){
                    grid.table(summary_df, rows= NULL)
                }else{
                    grid.table(summary_df[1:20,], rows= NULL)
                    mincount = 21
                    maxcount = 40
                    for (count20 in 1:(ceiling(nrow(summary_df)/20)-1)){
                        plot(0,type='n',axes=FALSE,ann=FALSE)
                        if (maxcount <= nrow(summary_df)){
                            grid.table(summary_df[mincount:maxcount,], rows= NULL)
                            mincount = mincount+20
                            maxcount = maxcount+20
                        }else{### stop last page with last sample otherwise NA in table
                            grid.table(summary_df[mincount:nrow(summary_df),], rows= NULL)}
                    }
                }

                ## optional output as .RData
                #if $output_rdata:
                    save(msidata.cv.pls, file="$classification_rdata")
                #end if


            ######################## PLS - analysis ###########################
            #elif str( $type_cond.method_cond.analysis_cond.PLS_method) == "PLS_analysis":
                print("PLS analysis")

                ## set variables for components and number of response groups
                component = c($type_cond.method_cond.analysis_cond.pls_comp)
                number_groups = length(levels(y_vector))

                ### stop if multiple values for PLS components are selected what sets component to 0
                tryCatch(
                        {

                        if (component==0)
                            {
                            stop(call.=FALSE)
                            }
                        },
                        error=function(cond) {
                        ## in case user used multiple inputs for component - this is only possible in cv apply
                            message("Error during PLS training")
                            message("Possible problems: Multiple values for component were selected - this is only possible in cvapply but not for PLS analysis or component was set to 0 but minimum for component is 1)")
                            stop(call.=FALSE)
                        }
                    )

                ### pls analysis and coefficients plot
                msidata.pls <- PLS(msidata, y = y_vector, ncomp = component, scale=$type_cond.method_cond.analysis_cond.pls_scale)
                plot(msidata.pls, main="PLS coefficients per m/z", col=colourvector)


                ## create new summary table
                summary_df = as.data.frame(summary(msidata.pls))
                colnames(summary_df) = c("Number of Components", "Accuracy", "Sensitivity", "Specificity")
                summary_df = round(summary_df, digits = 2)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                grid.table(summary_df, rows= NULL)

                ## Yweights plot: represent the importance of each response variable in predicting each component

                #if $type_cond.method_cond.analysis_cond.PLS_Yweights == "TRUE":
                Yweights = as.data.frame(msidata.pls@resultData@listData[[1]][["Yweights"]])
                Yweights = round(Yweights, digits = 4)
                Yweights.class <- cbind("class" = rownames(Yweights), Yweights)

                plot(0,type='n',axes=FALSE,ann=FALSE)
                text(x = 0.95, y = 1, "Yweights", cex = 2, font = 2)
                grid.table(Yweights.class, rows= NULL)

                #end if

                coefficient_plot = plot(msidata.pls, values="coefficients", lwd=2, main = "PLS coefficients per m/z")
                print(coefficient_plot)

                ## m/z and pixel information output
                pls_classes = data.frame(msidata.pls@resultData@listData[[1]][["class"]])

                ## pixel names and coordinates
                x_coords = msidata_coordinates@listData[["x"]]
                y_coords = msidata_coordinates@listData[["y"]]
                pixel_names = paste0("xy_", x_coords, "_", y_coords)

                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()

                pls_classes2 = data.frame(pixel_names, x_coords, y_coords, pls_classes, y_vector)
                colnames(pls_classes2) = c("pixel_name", "x", "y","predicted_class", "annotated_class")
                pls_classes2\$correct <- ifelse(pls_classes2\$predicted_class==pls_classes2\$annotated_class, T, F)

                write.table(pls_classes2, file="$pixeloutput", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

                correctness = round(sum(pls_classes2\$correct)/length(pls_classes2\$correct)*100,2)

                ## replace topFeatures table with coefficients table
                coefficients.df = as.data.frame(msidata.pls@resultData@listData[[1]][["coefficients"]])
                row_names <- msidata.pls@featureData@mz
                coefficients.df.rownames <- cbind("mz" = row_names, coefficients.df)
                write.table(coefficients.df.rownames, file = "$coefficients", quote = FALSE, sep = "\t", row.names = FALSE)

                ## add loadings and weights table
                loadings.df = as.data.frame(msidata.pls@resultData@listData[[1]][["loadings"]])
                loadings.df <- cbind("mz" = row_names, loadings.df)
                new_names <- paste0("loadings_", names(loadings.df)[-1])
                names(loadings.df)[-1] <- new_names

                weights.df = as.data.frame(msidata.pls@resultData@listData[[1]][["weights"]])
                weights.df <- cbind("mz" = row_names, weights.df)
                new_names <- paste0("weights_", names(weights.df)[-1])
                names(weights.df)[-1] <- new_names

                ## combine loading and weights table
                merged.load.wei = merge(loadings.df, weights.df, by = "mz")
                write.table(merged.load.wei, file = "$loadings_weights", quote = FALSE, sep = "\t", row.names = FALSE)

                ## image with predicted classes
                prediction_df = as.data.frame(cbind(coord(msidata.pls)[,1:2], pls_classes))
                colnames(prediction_df) = c("x", "y", "predicted_classes")

                prediction_plot = ggplot(prediction_df, aes(x=x, y=y, fill=predicted_classes))+
                       geom_tile() +
                       coord_fixed()+
                       ggtitle("Predicted condition for each pixel")+
			        theme_bw()+
		            theme(
		            plot.background = element_blank(),
		            panel.grid.major = element_blank(),
		            panel.grid.minor = element_blank())+
                       theme(text=element_text(family="ArialMT", face="bold", size=15))+
                       theme(legend.position="bottom",legend.direction="vertical")+
                       guides(fill=guide_legend(ncol=4,byrow=TRUE))+
                       scale_discrete_manual(aesthetics = c("colour", "fill"), values = colourvector)
                coord_labels = aggregate(cbind(x,y)~predicted_classes, data=prediction_df, mean, na.rm=TRUE, na.action="na.pass")
                ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$predicted_classes)
                print(prediction_plot)

	            ## correctness plot
	            correctness_plot = ggplot(pls_classes2, aes(x=x, y=y, fill=correct))+
                       geom_tile() +
                       coord_fixed()+
                       ggtitle(paste0("Correctness of classification: ", correctness, " %"))+
                       scale_fill_manual(values = c("TRUE" = "orange","FALSE" = "darkblue"))+
		                theme_bw()+
                        theme(
    		            plot.background = element_blank(),
   		                panel.grid.major = element_blank(),
  		                panel.grid.minor = element_blank())+
                        theme(text=element_text(family="ArialMT", face="bold", size=15))+
                        theme(legend.position="bottom",legend.direction="vertical")+
                        guides(fill=guide_legend(ncol=2,byrow=TRUE))
               coord_labels = aggregate(cbind(x,y)~correct, data=pls_classes2, mean, na.rm=TRUE, na.action="na.pass")
               ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$predicted_classes)
               print(correctness_plot)


                ### optional output as .RData
                #if $output_rdata:
                    save(msidata.pls, file="$classification_rdata")
                #end if

            #end if

        ######################## OPLS #############################
        #elif str( $type_cond.method_cond.class_method) == "OPLS":
            print("OPLS")

            ######################## OPLS -CV #############################
            #if str( $type_cond.method_cond.opls_analysis_cond.opls_method) == "opls_cvapply":
                print("OPLS cv")

                ## set variables for components and number of response groups
                components = c($type_cond.method_cond.opls_analysis_cond.opls_cvcomp)
                number_groups = length(levels(y_vector))

                ## OPLS-cvApply:
                msidata.cv.opls <- crossValidate(msidata, .y = y_vector, .fold = fold_vector, .fun = "OPLS", ncomp = components)

                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()



                ## new table with cv results to replace the old summary table
                results_list <- NULL
                for (i in seq_along(components)) {
                  ## extract accuracy, sensitivity, and specificity for the current i
                  accuracy <- round(as.data.frame(msidata.cv.opls@resultData@listData[[i]][["accuracy"]]), digits=2)
                  sensitivity <- round(as.data.frame(msidata.cv.opls@resultData@listData[[i]][["sensitivity"]]), digits=2)
                  specificity <- round(as.data.frame(msidata.cv.opls@resultData@listData[[i]][["specificity"]]), digits=2)

                  ## combine accuracy, sensitivity, and specificity into one data frame
                  result_df <- cbind(folds = rownames(accuracy), ncomp = i, accuracy, sensitivity, specificity)
                  colnames(result_df) <- c("folds", "ncomp", "accuracy", "sensitivity", "specificity")
                  rownames(result_df) <- NULL

                  ## add column names with ncomp as first row to each dataframe
                  col_names_row <- data.frame(folds = "folds", ncomp = paste0("ncomp", i), accuracy = "accuracy", sensitivity = "sensitivity", specificity = "specificity")
                  result_df <- rbind(col_names_row, result_df)

                  results_list[[i]] <- result_df
                }

                ## combine all data frames in the list into one data frame
                results_df <- do.call(rbind, results_list)

                summary_df <- results_df

                ## new table and plot of accuracies over all components
                summary.cv.opls = as.data.frame(summary(msidata.cv.opls))

                ## table with values
                plot(0,type='n',axes=FALSE,ann=FALSE)
                summary.cv.opls.round <- round(summary.cv.opls, digits=2)
                grid.table(summary.cv.opls.round, rows=NULL)

                accuracy_plot = ggplot(summary.cv.opls, aes(x = ncomp, y = Accuracy)) +
                  geom_point(color = "blue", size = 3) +   # Add points
                  geom_line() +
                  theme_bw()
                print(accuracy_plot)

                ## print table with summary in pdf
                par(opar)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                title(main="Summary for the different components\n", adj=0.5)
                ## 20 rows fits in one page:
                if (nrow(summary_df)<=20){
                    grid.table(summary_df, rows= NULL)
                }else{
                    grid.table(summary_df[1:20,], rows= NULL)
                    mincount = 21
                    maxcount = 40
                    for (count20 in 1:(ceiling(nrow(summary_df)/20)-1)){
                        plot(0,type='n',axes=FALSE,ann=FALSE)
                        if (maxcount <= nrow(summary_df)){
                            grid.table(summary_df[mincount:maxcount,], rows= NULL)
                            mincount = mincount+20
                            maxcount = maxcount+20
                        }else{### stop last page with last sample otherwise NA in table
                            grid.table(summary_df[mincount:nrow(summary_df),], rows= NULL)}
                    }
                }

                ## optional output as .RData
                #if $output_rdata:
                    save(msidata.cv.opls, file="$classification_rdata")
                #end if


            ######################## OPLS -analysis ###########################
            #elif str( $type_cond.method_cond.opls_analysis_cond.opls_method) == "opls_analysis":
                print("OPLS analysis")

                ## set variables for components and number of response groups
                component = c($type_cond.method_cond.opls_analysis_cond.opls_comp)
                number_groups = length(levels(y_vector))

                ### stop if multiple values for OPLS components are selected what sets component to 0
                tryCatch(
                        {

                        if (component==0)
                            {
                            stop(call.=FALSE)
                            }
                        },
                        error=function(cond) {
                        ## in case user used multiple inputs for component - this is only possible in cv apply
                            message("Error during OPLS training")
                            message("Possible problems: Multiple values for component were selected - this is only possible in cvapply but not for OPLS analysis or component was set to 0 but minimum for component is 1)")
                            stop(call.=FALSE)
                        }
                    )

                ### opls analysis and coefficients plot
                msidata.opls <- OPLS(msidata, y = y_vector, ncomp = component, scale=$type_cond.method_cond.opls_analysis_cond.opls_scale)
                plot(msidata.opls, main="OPLS coefficients per m/z", col=colourvector)

                ## create new summary table
                summary_df = as.data.frame(summary(msidata.opls))
                colnames(summary_df) = c("Number of Components", "Accuracy", "Sensitivity", "Specificity")
                summary_df = round(summary_df, digits = 2)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                grid.table(summary_df, rows= NULL)


                #if $type_cond.method_cond.opls_analysis_cond.OPLS_Yweights == "TRUE":
                ## Yweights plot: represent the importance of each response variable in predicting each component
                Yweights = as.data.frame(msidata.opls@resultData@listData[[1]][["Yweights"]])
                Yweights = round(Yweights, digits = 4)
                Yweights.class <- cbind("class" = rownames(Yweights), Yweights)

                plot(0,type='n',axes=FALSE,ann=FALSE)
                text(x = 0.95, y = 1, "Yweights", cex = 2, font = 2)
                grid.table(Yweights.class, rows= NULL)
                #end if

                coefficient_plot = plot(msidata.opls, values="coefficients", lwd=2, main = "OPLS coefficients per m/z")
                print(coefficient_plot)

                ## m/z and pixel information output
                opls_classes = data.frame(msidata.opls@resultData@listData[[1]][["class"]])

                ## pixel names and coordinates
                x_coords = msidata_coordinates@listData[["x"]]
                y_coords = msidata_coordinates@listData[["y"]]
                pixel_names = paste0("xy_", x_coords, "_", y_coords)

                opls_classes2 = data.frame(pixel_names, x_coords, y_coords, opls_classes, y_vector)
                colnames(opls_classes2) = c("pixel names", "x", "y","predicted_class", "annotated_class")
                opls_classes2\$correct <- ifelse(opls_classes2\$predicted_class == opls_classes2\$annotated_class, T, F)

                write.table(opls_classes2, file="$pixeloutput", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

                correctness = round(sum(opls_classes2\$correct)/length(opls_classes2\$correct)*100,2)

                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()

                ## replace topFeatures table with coefficients table
                coefficients.df = as.data.frame(msidata.opls@resultData@listData[[1]][["coefficients"]])
                row_names <- msidata.opls@featureData@mz
                coefficients.df.rownames <- cbind("mz" = row_names, coefficients.df)
                write.table(coefficients.df.rownames, file = "$coefficients", quote = FALSE, sep = "\t", row.names = FALSE)

                ## add loadings and weights table
                loadings.df = as.data.frame(msidata.opls@resultData@listData[[1]][["loadings"]])
                loadings.df <- cbind("mz" = row_names, loadings.df)
                new_names <- paste0("loadings_", names(loadings.df)[-1])
                names(loadings.df)[-1] <- new_names

                weights.df = as.data.frame(msidata.opls@resultData@listData[[1]][["weights"]])
                weights.df <- cbind("mz" = row_names, weights.df)
                new_names <- paste0("weights_", names(weights.df)[-1])
                names(weights.df)[-1] <- new_names

                ## combine loading and weights table
                merged.load.wei = merge(loadings.df, weights.df, by = "mz")
                write.table(merged.load.wei, file = "$loadings_weights", quote = FALSE, sep = "\t", row.names = FALSE)

                ## image with predicted classes
                prediction_df = as.data.frame(cbind(coord(msidata.opls)[,1:2], opls_classes))
                colnames(prediction_df) = c("x", "y", "predicted_classes")

                prediction_plot = ggplot(prediction_df, aes(x=x, y=y, fill=predicted_classes))+
                       geom_tile() +
                       coord_fixed()+
                       ggtitle("Predicted condition for each pixel")+
			            theme_bw()+
			            theme(
		                plot.background = element_blank(),
		                panel.grid.major = element_blank(),
		                panel.grid.minor = element_blank())+
                       theme(text=element_text(family="ArialMT", face="bold", size=15))+
                       theme(legend.position="bottom",legend.direction="vertical")+
                       guides(fill=guide_legend(ncol=4,byrow=TRUE))+
                       scale_discrete_manual(aesthetics = c("colour", "fill"), values = colourvector)
                coord_labels = aggregate(cbind(x,y)~predicted_classes, data=prediction_df, mean, na.rm=TRUE, na.action="na.pass")
                ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$predicted_classes)
                print(prediction_plot)

                ## correctness plot
                 correctness_plot = ggplot(opls_classes2, aes(x=x, y=y, fill=correct))+
                       geom_tile() +
                       coord_fixed()+
                       ggtitle(paste0("Correctness of classification: ", correctness, " %"))+
                       scale_fill_manual(values = c("TRUE" = "orange","FALSE" = "darkblue"))+
		                theme_bw()+
                        theme(
    		            plot.background = element_blank(),
   		                panel.grid.major = element_blank(),
  		                panel.grid.minor = element_blank())+
                        theme(text=element_text(family="ArialMT", face="bold", size=15))+
                        theme(legend.position="bottom",legend.direction="vertical")+
                        guides(fill=guide_legend(ncol=2,byrow=TRUE))
                coord_labels = aggregate(cbind(x,y)~correct, data=opls_classes2, mean, na.rm=TRUE, na.action="na.pass")
                ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$predicted_classes)
                print(correctness_plot)


                 ## optional output as .RData
                 #if $output_rdata:
                    save(msidata.opls, file="$classification_rdata")
                 #end if
            #end if


        ######################## SSC #############################
        #elif str( $type_cond.method_cond.class_method) == "spatialShrunkenCentroids":
            print("SSC")

            ######################## SSC - CV #############################
            #if str( $type_cond.method_cond.ssc_analysis_cond.ssc_method) == "ssc_cvapply":
                print("SSC cv")

                ## set variables for components and number of response groups
                number_groups = length(levels(y_vector))

                ## SSC-cvApply:
                msidata.cv.ssc <- crossValidate(msidata, .y = y_vector,.fold = fold_vector,.fun = "spatialShrunkenCentroids", r = c($type_cond.method_cond.ssc_r), s = c($type_cond.method_cond.ssc_s), method = "$type_cond.method_cond.ssc_kernel_method")


                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()

                ## new table and plot of accuracies over all components
                summary.cv.ssc = as.data.frame(summary(msidata.cv.ssc))
                summary.cv.ssc.round <- round(summary.cv.ssc, digits=2)

                par(opar)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                title(main="Summary for the different parameters\n", adj=0.5)
                ## 20 rows fits in one page:
                if (nrow(summary.cv.ssc.round)<=20){
                    grid.table(summary.cv.ssc.round, rows= NULL)
                }else{
                    grid.table(summary.cv.ssc.round[1:20,], rows= NULL)
                    mincount = 21
                    maxcount = 40
                    for (count20 in 1:(ceiling(nrow(summary.cv.ssc.round)/20)-1)){
                        plot(0,type='n',axes=FALSE,ann=FALSE)
                        if (maxcount <= nrow(summary.cv.ssc.round)){
                            grid.table(summary.cv.ssc.round[mincount:maxcount,], rows= NULL)
                            mincount = mincount+20
                            maxcount = maxcount+20
                        }else{### stop last page with last sample otherwise NA in table
                            grid.table(summary.cv.ssc.round[mincount:nrow(summary.cv.ssc.round),], rows= NULL)}
                    }
                }

                ## new accuracy plots
                #if $type_cond.method_cond.ssc_analysis_cond.ssc_cv_accuracy_plot == "TRUE":
                accuracy_plot = ggplot(summary.cv.ssc, aes(x = s, y = Accuracy)) +
                  geom_point(color = "blue", size = 3) +   # Add points
                  geom_line() +
                  theme_bw() +
                  facet_wrap(~ r)

                print(accuracy_plot)

                ## or as alternative accuracy plot for each r value on own page:
                #elif $type_cond.method_cond.ssc_analysis_cond.ssc_cv_accuracy_plot == "FALSE":
                unique_r_values <- unique(summary.cv.ssc\$r)

                for (r_value in unique_r_values) {
                  ## Create a subset for the current value of r
                  plot_data <- subset(summary.cv.ssc, r == r_value)
                  ## Create the accuracy plot for the current value of r
                  accuracy_plot <- ggplot(plot_data, aes(x = s, y = Accuracy)) +
                                     geom_point(color = "blue", size = 3) +   # Add points
                                     geom_line() +
                                     theme_bw() +
                                     ggtitle(paste("Plot for r =", r_value)) +  # Add a title
                                     theme(plot.title = element_text(hjust = 0.5))  # Center the title
                  print(accuracy_plot)
                }
                #end if

                ## table with cv values per fold group for each combination of r and s
                r_s_df = as.data.frame(msidata.cv.ssc@modelData@listData)
                r_s_df\$parameter = paste0("r=", r_s_df\$r, " and s=", r_s_df\$s)
                iteration = seq_along(r_s_df\$parameter)

                results_list <- NULL
                for (i in iteration) {
                    ## extract accuracy, sensitivity, and specificity for the current i
                    accuracy <- round(as.data.frame(msidata.cv.ssc@resultData@listData[[i]][["accuracy"]]), digits=2)
                    sensitivity <- round(as.data.frame(msidata.cv.ssc@resultData@listData[[i]][["sensitivity"]]), digits=2)
                    specificity <- round(as.data.frame(msidata.cv.ssc@resultData@listData[[i]][["specificity"]]), digits=2)

                    ## combine accuracy, sensitivity, and specificity into one data frame
                    result_df <- cbind(folds = rownames(accuracy), parameter = r_s_df\$parameter[i], accuracy, sensitivity, specificity)
                    colnames(result_df) <- c("folds", "parameter", "accuracy", "sensitivity", "specificity")
                    rownames(result_df) <- NULL

                    ## add column names as first row to each dataframe
                    col_names_row <- data.frame(folds = "folds", parameter = "parameter", accuracy = "accuracy", sensitivity = "sensitivity", specificity = "specificity")
                    result_df <- rbind(col_names_row, result_df)

                    results_list[[i]] <- result_df
                  }

                ## combine all data frames in the list into one data frame
                results_df <- do.call(rbind, results_list)
                summary_df <- results_df

                par(opar)
                plot(0,type='n',axes=FALSE,ann=FALSE)
                title(main="More advanced folds output table: \n Summary for each fold\n", adj=0.5)
                ## 20 rows fits in one page:
                if (nrow(summary_df)<=20){
                    grid.table(summary_df, rows= NULL)
                }else{
                    grid.table(summary_df[1:20,], rows= NULL)
                    mincount = 21
                    maxcount = 40
                    for (count20 in 1:(ceiling(nrow(summary_df)/20)-1)){
                        plot(0,type='n',axes=FALSE,ann=FALSE)
                        if (maxcount <= nrow(summary_df)){
                            grid.table(summary_df[mincount:maxcount,], rows= NULL)
                            mincount = mincount+20
                            maxcount = maxcount+20
                        }else{### stop last page with last sample otherwise NA in table
                            grid.table(summary_df[mincount:nrow(summary_df),], rows= NULL)}
                    }
                }


                ## new code to extract best r and s values
                max_accuracy_index <- which.max(summary.cv.ssc\$Accuracy)

                ## extract the corresponding values of "r" and "s"
                highest_accuracy_r <- summary.cv.ssc\$r[max_accuracy_index]
                highest_accuracy_s <- summary.cv.ssc\$s[max_accuracy_index]

		        #if $type_cond.method_cond.ssc_analysis_cond.write_best_params:
                	write.table(highest_accuracy_r, file="$best_r", quote = FALSE, row.names = FALSE, col.names=FALSE, sep = "\t")
                	write.table(highest_accuracy_s, file="$best_s", quote = FALSE, row.names = FALSE, col.names=FALSE, sep = "\t")
                #end if

                ## optional output as .RData
                #if $output_rdata:
                    save(msidata.cv.ssc, file="$classification_rdata")
                #end if

            ######################## SSC -analysis ###########################
            #elif str( $type_cond.method_cond.ssc_analysis_cond.ssc_method) == "ssc_analysis":
                print("SSC analysis")

                ## set variables for components and number of response groups
                number_groups = length(levels(y_vector))

                ## SSC analysis and plot
                msidata.ssc <- spatialShrunkenCentroids(msidata, y = y_vector, r = c($type_cond.method_cond.ssc_r), s = c($type_cond.method_cond.ssc_s), method = "$type_cond.method_cond.ssc_kernel_method")
                print(plot(msidata.ssc, values = "statistic", model = list(r = c($type_cond.method_cond.ssc_r), s = c($type_cond.method_cond.ssc_s)), col=colourvector, lwd=2))

                ### stop if multiple values for r and s were used as input
                tryCatch(
                        {

                        if (length(names(msidata.ssc@resultData))>1)
                            {
                            stop(call.=FALSE)
                            }
                        },
                        error=function(cond) {
                        ## in case user used multiple inputs for r or s stop - this is only possible in cv apply
                            message("Error during SSC training")
                            message("Possible problem: multiple values for r or s selected - this is only possible in cvapply but not for spatial shrunken centroid analysis)")
                            stop(call.=FALSE)
                        }
                    )

                summary_df = as.data.frame(summary(msidata.ssc))
                summary_df = round(summary_df, digits=3)
                colnames(summary_df) = c("Radius r", "Shrinkage s", "Features/Class", "Accuracy", "Sensitivity", "Specificity")
                plot(0,type='n',axes=FALSE,ann=FALSE)
                grid.table(summary_df, rows= NULL)

                ## image of the best m/z
                minimumy = min(coord(msidata)[,2])
                maximumy = max(coord(msidata)[,2])
                print(image(msidata, mz = topFeatures(msidata.ssc)[1,1], normalize.image = "linear", contrast.enhance = "histogram",smooth.image="gaussian", ylim= c(maximumy+0.2*maximumy,minimumy-0.2*minimumy), main="best m/z heatmap"))

                ## m/z and pixel information output
                x_coords = msidata_coordinates@listData[["x"]]
                y_coords = msidata_coordinates@listData[["y"]]
                pixel_names = paste0("xy_", x_coords, "_", y_coords)


                ## remove msidata to clean up RAM space
                rm(msidata)
                gc()


                ## toplabel (m/z features output)
                ssc_toplabels = topFeatures(msidata.ssc, n=$type_cond.method_cond.ssc_toplabels)
                ssc_toplabels@listData[["centers"]] = round (ssc_toplabels@listData[["centers"]], digits = 6)
                ssc_toplabels@listData[["statistic"]] = round (ssc_toplabels@listData[["statistic"]], digits = 6)
                write.table(ssc_toplabels, file="$mzfeatures", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

                print(image(msidata.ssc, model=list(r = c($type_cond.method_cond.ssc_r), s = c($type_cond.method_cond.ssc_s)), ylim= c(maximumy+0.2*maximumy,minimumy-0.2*minimumy), col=colourvector, values="class", layout=c(1,1), main="Class Prediction"))
                print(image(msidata.ssc, model=list(r = c($type_cond.method_cond.ssc_r), s = c($type_cond.method_cond.ssc_s)), ylim= c(maximumy+0.2*maximumy,minimumy-0.2*minimumy), col=colourvector, values="probability", layout=c(1,1), main="Class Probabilities"))


                ## pixel output with correctness
                ssc_classes = data.frame(msidata.ssc@resultData@listData[[1]][["class"]])
                colnames(ssc_classes) = "predicted_class"
                ssc_classes\$predicted_class = ifelse(is.na(ssc_classes\$predicted_class), "NA", as.character(ssc_classes\$predicted_class))
                ssc_probabilities = data.frame(msidata.ssc@resultData@listData[[1]][["probability"]])


                ssc_classes2 = data.frame(pixel_names, x_coords, y_coords, ssc_classes, ssc_probabilities, y_vector)
                colnames(ssc_classes2) = c("pixel_names", "x", "y","predicted_classes", levels(msidata.ssc@resultData@listData[[1]][["class"]]), "annotated_class")
                ssc_classes2\$correct<- ifelse(ssc_classes2\$predicted_classes==ssc_classes2\$annotated_class, T, F)
	            correctness = round(sum(ssc_classes2\$correct, na.rm = TRUE)/length(ssc_classes2\$correct)*100,2)

                write.table(ssc_classes2, file="$pixeloutput", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

	            correctness_plot = ggplot(ssc_classes2, aes(x=x, y=y, fill=correct))+
                       geom_tile() +
                       coord_fixed()+
                       ggtitle(paste0("Correctness of classification: ", correctness, " %"))+
                       scale_fill_manual(values = c("TRUE" = "orange","FALSE" = "darkblue"))+
		                theme_bw()+
                        theme(
    		            plot.background = element_blank(),
   		                panel.grid.major = element_blank(),
  		                panel.grid.minor = element_blank())+
                        theme(text=element_text(family="ArialMT", face="bold", size=15))+
                        theme(legend.position="bottom",legend.direction="vertical")+
                        guides(fill=guide_legend(ncol=2,byrow=TRUE))
                coord_labels = aggregate(cbind(x,y)~correct, data=ssc_classes2, mean, na.rm=TRUE, na.action="na.pass")
                ##coord_labels\$file_number = gsub( "_.*$", "", coord_labels\$predicted_classes)
                print(correctness_plot)


                ## optional output as .RData
                #if $output_rdata:
                    save(msidata.ssc, file="$classification_rdata")
                #end if
            #end if
        #end if



    ######################## II) Prediction #############################
    #############################################################################

    #elif str($type_cond.type_method) == "prediction":
        print("prediction")

        training_data = loadRData("$type_cond.training_result")

        #if str($type_cond.new_y_values_cond.new_y_values) == "new_response":
            print("new response")

            new_y_tabular = read.delim("$type_cond.new_y_values_cond.new_response_file", header = $type_cond.new_y_values_cond.new_tabular_header, stringsAsFactors = FALSE)
            new_y_input = new_y_tabular[,c($type_cond.new_y_values_cond.column_new_x, $type_cond.new_y_values_cond.column_new_y, $type_cond.new_y_values_cond.column_new_response)]
            colnames(new_y_input)[1:2] = c("x", "y")

            ## merge with coordinate information of msidata
            msidata_coordinates = cbind(coord(msidata)[,1:2], c(1:ncol(msidata)))
            colnames(msidata_coordinates)[3] = "pixel_index"
            merged_response = as.data.frame(merge(msidata_coordinates, new_y_input, by=c("x", "y"), all.x=TRUE))
            merged_response[is.na(merged_response)] = "NA"
            merged_response = merged_response[order(merged_response\$pixel_index),]
            new_y_vector = as.factor(merged_response[,4])

            prediction = predict(training_data, msidata, newy = new_y_vector)

            ##numbers of levels for colour selection
            number_levels = length(levels(new_y_vector))

	        ##new summary table

	        ##if SSC classification, summary table has more results:
	        #if str($type_cond.classification_type) == "SSC_classifier":
            print("SSC classification summary")

            summary_df = as.data.frame(summary(prediction))
            summary_df = round(summary_df, digits=3)
            colnames(summary_df) = c("Radius r", "Shrinkage s", "Features/Class", "Accuracy", "Sensitivity", "Specificity")
            plot(0,type='n',axes=FALSE,ann=FALSE)
            grid.table(summary_df, rows= NULL)

	        ## else PLS or OPLS classifier:
            #else
            print("PLS/OPLS classifier")
            summary_df = as.data.frame(summary(prediction))
            colnames(summary_df) = c("Component", "Accuracy", "Sensitivity", "Specificity")
            summary_df = round(summary_df, digits = 2)
            plot(0,type='n',axes=FALSE,ann=FALSE)
            grid.table(summary_df, rows= NULL)

            #end if


        ##else for prediction without a new annotation (no calculation of accuracy):
        #else
            prediction = predict(training_data, msidata)
            number_levels = length(levels(training_data@resultData@listData[[1]][["class"]]))
        #end if

        ## colours selection:

	    #if str($colour_conditional.colour_type) == "manual_colour"
	    #set $color_string = ','.join(['"%s"' % $color.annotation_color for $color in $colour_conditional.colours])
	        colourvector = c($color_string)

	    #elif str($colour_conditional.colour_type) == "colourpalette"
	    colourvector = noquote($colour_conditional.palettes)(number_levels)

	    #end if

        ## m/z and pixel information output
        predicted_classes = data.frame(prediction@resultData@listData[[1]][["class"]])
        msidata_coordinates = cbind(coord(msidata)[,1:2], c(1:ncol(msidata)))
        colnames(msidata_coordinates)[3] = "pixel_index"
        x_coords = msidata_coordinates@listData[["x"]]
        y_coords = msidata_coordinates@listData[["y"]]
        pixel_names = paste0("xy_", x_coords, "_", y_coords)
        predicted_classes2 = data.frame(pixel_names, x_coords, y_coords, predicted_classes)
        colnames(predicted_classes2) = c("pixel names", "x", "y","predicted condition")

        ##topFeatures only available for SSC; for PLS and OPLS coefficients loading and weights are provided

        #if str($type_cond.classification_type) == "SSC_classifier":
        predicted_toplabels = topFeatures(prediction, n=$type_cond.classification_type_cond.predicted_toplabels)
        predicted_toplabels <- as.data.frame(predicted_toplabels)
        predicted_toplabels[,6:7] <-round(predicted_toplabels[,6:7], digits = 5)
        write.table(predicted_toplabels, file="$mzfeatures", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

        #else
	    ## if PLS or OPLS classifier, coefficients, loadings, and weights instead of topFeatures
	    coefficients.df = as.data.frame(prediction@resultData@listData[[1]][["coefficients"]])
        row_names <- prediction@featureData@mz
        coefficients.df <- cbind("mz" = row_names, coefficients.df)
        write.table(coefficients.df, file = "$coefficients", quote = FALSE, sep = "\t", row.names = FALSE)

        ## add loadings and weights table
        loadings.df = as.data.frame(prediction@resultData@listData[[1]][["loadings"]])
        loadings.df <- cbind("mz" = row_names, loadings.df)
        new_names <- paste0("loadings_", names(loadings.df)[-1])
        names(loadings.df)[-1] <- new_names

        weights.df = as.data.frame(prediction@resultData@listData[[1]][["weights"]])
        weights.df <- cbind("mz" = row_names, weights.df)
        new_names <- paste0("weights_", names(weights.df)[-1])
        names(weights.df)[-1] <- new_names

        ## combine loading and weights table
        merged.load.wei = merge(loadings.df, weights.df, by = "mz")
        write.table(merged.load.wei, file = "$loadings_weights", quote = FALSE, sep = "\t", row.names = FALSE)

        #end if

        ##predicted classes
        prediction_df = as.data.frame(cbind(coord(prediction)[,1:2], predicted_classes))
        colnames(prediction_df) = c("x", "y", "predicted_classes")

        #if str($type_cond.classification_type) == "SSC_classifier":
            ## this seems to work only for SSC, therefore overwrite tables
            predicted_probabilities = data.frame(prediction@resultData@listData[[1]][["probability"]])
            predicted_classes2 = data.frame(pixel_names, x_coords, y_coords, predicted_classes, predicted_probabilities)
            colnames(predicted_classes2) = c("pixel names", "x", "y","predicted condition", levels(prediction@resultData@listData[[1]][["class"]]))
            ## also image modes are specific to SSC
            print(predicted_classes2[1:5,])
            print(image(prediction, values="class", layout=c(1,1), main="Class Prediction", ylim= c(maximumy+0.2*maximumy,minimumy-0.2*minimumy), col=colourvector))
            print(image(prediction, values="probability", layout=c(1,1), main="Class Probabilities",ylim= c(maximumy+0.2*maximumy,minimumy-0.2*minimumy), col=colourvector))

	    #else
            prediction_plot = ggplot(prediction_df, aes(x=x, y=y, fill=predicted_classes))+
        	geom_tile()+
        	coord_fixed()+
        	ggtitle("Predicted condition for each spectrum")+
        	theme_bw()+
        	theme(
        	plot.background = element_blank(),
        	panel.grid.major = element_blank(),
        	panel.grid.minor = element_blank())+
        	theme(text=element_text(family="ArialMT", face="bold", size=15))+
        	theme(legend.position="bottom", legend.direction="vertical")+
        	guides(fill=guide_legend(ncol=4, byrow=TRUE))+
        	scale_discrete_manual(aesthetics = c("colour", "fill"), values = colourvector)
        	coord_labels = aggregate(cbind(x,y)~predicted_classes, data=prediction_df, mean, na.rm=TRUE, na.action="na.pass")
        	##coord_labels\$file_number = gsub( "_.*§", "", coord_labels\$predicted_classes)
        	print(prediction_plot)
        #end if

        #if str($type_cond.new_y_values_cond.new_y_values) == "new_response":
        ## image with right and wrong classes:
	    comparison_df = as.data.frame(cbind(prediction_df, new_y_vector))
	    colnames(comparison_df) = c("x", "y", "predicted_class", "annotated_class")
	    comparison_df\$predicted_class = ifelse(is.na(comparison_df\$predicted_class), "NA", as.character(comparison_df\$predicted_class))
        comparison_df\$correct <- ifelse(comparison_df\$predicted_class==comparison_df\$annotated_class, T, F)
	    correctness = round(sum(comparison_df\$correct, na.rm = TRUE)/length(comparison_df\$correct)*100,2)

	    correctness_plot = ggplot(comparison_df, aes(x=x, y=y, fill=correct))+
                       geom_tile()+
                       scale_fill_manual(values = c("TRUE" = "orange","FALSE" = "darkblue"))+
                       coord_fixed()+
                       ggtitle(paste0("Correctness of classification: ", correctness, " %"))+
                       theme_bw()+
                       theme(text=element_text(family="ArialMT", face="bold", size=15))+
                       theme(legend.position="bottom",legend.direction="vertical")+
                       guides(fill=guide_legend(ncol=2,byrow=TRUE))
        print(correctness_plot)
        #end if

        ## pixel output
        #if str($type_cond.new_y_values_cond.new_y_values) == "new_response":
            print("new response output")
            write.table(comparison_df, file="$pixeloutput", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

        #else
            write.table(predicted_classes2, file="$pixeloutput", quote = FALSE, row.names = FALSE, col.names=TRUE, sep = "\t")

        #end if

        ## optional output as .RData
        #if $output_rdata:
            msidata = prediction
            save(msidata, file="$classification_rdata")
        #end if

    #end if

    dev.off()

}else{
    plot.new()
    text(0.5, 0.5, "Inputfile has no intensities > 0  \n or contains NA values.", cex = 1.5)
    print("Inputfile has no intensities > 0 or contains NA values")
    dev.off()
}


    ]]></configfile>
    </configfiles>
    <inputs>
        <expand macro="reading_msidata"/>
        <conditional name="type_cond">
            <param name="type_method" type="select" label="Analysis step to perform">
                <option value="training" selected="True">training</option>
                <option value="prediction">prediction</option>
            </param>
            <when value="training">

                <param name="annotation_file" type="data" format="tabular" label="Load tabular file with pixel coordinates and their classes"
                help="Three or four columns: x values, y values, response values, optionally fold values"/>
                <param name="column_x" data_ref="annotation_file" label="Column with x values" type="data_column"/>
                <param name="column_y" data_ref="annotation_file" label="Column with y values" type="data_column"/>
                <param name="column_response" data_ref="annotation_file" label="Column with response (condition) values" type="data_column" help="This is the condition (pixel group) which will be classified"/>
                <param name="column_fold" data_ref="annotation_file" optional="True" label="Column with fold values - only neccessary for cvapply" type="data_column" help="Each fold must contain pixels of all response groups and is used for cross validation"/>
                <param name="tabular_header" type="boolean" label="Tabular files contain a header line" truevalue="TRUE" falsevalue="FALSE"/>

                <conditional name="method_cond">
                    <param name="class_method" type="select" label="Select the method for classification">
                        <option value="PLS" selected="True">PLS-DA</option>
                        <option value="OPLS">OPLS-DA</option>
                        <option value="spatialShrunkenCentroids">spatial shrunken centroids</option>
                    </param>
                    <when value="PLS">

                        <conditional name="analysis_cond">
                            <param name="PLS_method" type="select" label="Crossvalidation or analysis">
                                <option value="cvapply" selected="True">cvApply</option>
                                <option value="PLS_analysis">PLS-DA analysis</option>
                            </param>
                            <when value="cvapply">
                                <param name="plscv_comp" type="text" value="1:2"
                                       label="The number of PLS-DA components" help="For cvapply multiple values are allowed (e.g. 1,2,3 or 2:5). Mininum is 1.">
                                <expand macro="sanitizer_multiple_digits"/>
                                </param>
                            </when>
                            <when value="PLS_analysis">
                                <param name="pls_comp" type="integer" value="5"
                                       label="The optimal number of PLS-DA components as indicated by cross-validations (minimum is 1)" help="Run cvApply first to optain optimal number of PLS-DA components"/>
                                <param name="pls_scale" type="boolean" label="Data scaling" truevalue="TRUE" falsevalue="FALSE"/>
                                <param name="PLS_Yweights" type="boolean" label="Y weights" help="Y weights represent the coefficients associated with the response variables and are used to model the relationship between predictors and responses in the context of classification. They represent the importance of each response variable in predicting each component. They can be useful if you have multiple response variables."/>
                                <!--param name="pls_toplabels" type="integer" value="100
                                   label="Number of toplabels (m/z features) which should be written in tabular output"/-->
                            </when>
                        </conditional>
                    </when>

                    <when value="OPLS">

                        <conditional name="opls_analysis_cond">
                            <param name="opls_method" type="select" label="Analysis step to perform">
                                <option value="opls_cvapply" selected="True">cvApply</option>
                                <option value="opls_analysis">OPLS-DA analysis</option>
                            </param>

                            <when value="opls_cvapply">
                                <param name="opls_cvcomp" type="text" value="1:2"
                                       label="The number of OPLS-DA components" help="For cvapply multiple values are allowed (e.g. 1,2,3 or 2:5). Minimum is 1.">
                                <expand macro="sanitizer_multiple_digits"/>
                                </param>
                                <!--param name="xnew_cv" type="boolean" truevalue="TRUE" falsevalue="FALSE" label="Keep new matrix"/-->
                            </when>

                            <when value="opls_analysis">
                                <param name="opls_comp" type="integer" value="5"
                                       label="The optimal number of OPLS-DA components as indicated by cross-validations (minimum is 1)" help="Run cvApply first to optain optimal number of OPLS-DA components"/>
                                <!--param name="xnew" type="boolean" truevalue="TRUE" falsevalue="FALSE" label="Keep new matrix"/-->
                                <param name="opls_scale" type="boolean" truevalue="TRUE" falsevalue="FALSE" label="Data scaling"/>
                                <param name="OPLS_Yweights" type="boolean" label="Y weights" help="Y weights represent the coefficients associated with the response variables and are used to model the relationship between predictors and responses in the context of classification. They represent the importance of each response variable in predicting each component. They can be useful if you have multiple response variables."/>
                                <!--param name="opls_toplabels" type="integer" value="100"
                                   label="Number of toplabels (m/z features) which should be written in tabular output"/-->
                            </when>
                        </conditional>
                    </when>

                    <when value="spatialShrunkenCentroids">
                        <conditional name="ssc_analysis_cond">
                            <param name="ssc_method" type="select" label="Analysis step to perform">
                                <option value="ssc_cvapply" selected="True">cvApply</option>
                                <option value="ssc_analysis">spatial shrunken centroids analysis</option>
                            </param>
                            <when value="ssc_cvapply">
                                  <param name="write_best_params" type="boolean" label="Write out best r and s values" help="Can be used to generate automatic classification workflow"/>
                                  <param name="ssc_cv_accuracy_plot" type="boolean" label="Plot CV accuracy plots on one page (=Yes) or individual pages (=No)"/>
                            </when>
                            <when value="ssc_analysis">
                                <param name="ssc_toplabels" type="integer" value="100"
                                   label="Number of toplabels (m/z features) which should be written in tabular output"/>
                            </when>
                        </conditional>
                        <param name="ssc_r" type="text" value="2"
                               label="The spatial neighborhood radius of nearby pixels to consider (r)" help="For cvapply multiple values are allowed (e.g. 0,1,2,3 or 2:5)">
                        <expand macro="sanitizer_multiple_digits"/>
                        </param>
                        <param name="ssc_s" type="text" value="2"
                               label="The sparsity thresholding parameter by which to shrink the t-statistics (s)." help="For cvapply multiple values are allowed (e.g. 0,1,2 or 2:5)">
                        <expand macro="sanitizer_multiple_digits"/>
                        </param>
                        <param name="ssc_kernel_method" type="select" display="radio" label = "The method to use to calculate the spatial smoothing kernels for the embedding. The 'gaussian' method refers to spatially-aware (SA) weights, and 'adaptive' refers to spatially-aware structurally-adaptive (SASA) weights">
                            <option value="gaussian">gaussian</option>
                            <option value="adaptive" selected="True">adaptive</option>
                        </param>
                    </when>
                </conditional>

            </when>

            <when value="prediction">
                <param name="training_result" type="data" format="rdata" label="Result from previous classification training"/>
                <conditional name="classification_type_cond">
                    <param name="classification_type" type="select" label="Which classification method was used">
                	    <option value="PLS_classifier" selected="True" >PLS classifier</option>
                	    <option value="OPLS_classifier">OPLS classifier</option>
                	    <option value="SSC_classifier">SSC classifier</option>
                    </param>
                    <when value="PLS_classifier"/>
                    <when value="OPLS_classifier"/>
                    <when value="SSC_classifier">
                        <param name="predicted_toplabels" type="integer" value="100"
                                   label="Number of toplabels (m/z features) which should be written in tabular output"/>
                    </when>
                </conditional>
                <conditional name="new_y_values_cond">
                    <param name="new_y_values" type="select" label="Load annotations (optional, but allows accuracy calculations)">
                        <option value="no_new_response" selected="True">no</option>
                        <option value="new_response">use annotations</option>
                    </param>
                    <when value="no_new_response"/>
                    <when value="new_response">
                        <param name="new_response_file" type="data" format="tabular" label="Load tabular file with pixel coordinates and the new response"/>
                        <param name="column_new_x" data_ref="new_response_file" label="Column with x values" type="data_column"/>
                        <param name="column_new_y" data_ref="new_response_file" label="Column with y values" type="data_column"/>
                        <param name="column_new_response" data_ref="new_response_file" label="Column with new response values" type="data_column"/>
                        <param name="new_tabular_header" type="boolean" label="Tabular files contain a header line" truevalue="TRUE" falsevalue="FALSE"/>
                    </when>
                </conditional>
            </when>
        </conditional>
        <conditional name="colour_conditional">
	    <param name="colour_type" type="select" label="Choose a colour scheme">
	        <option value="colourpalette" selected="True" >Colour palette</option>
	        <option value="manual_colour">Manual selection</option>
	    </param>
	    <when value="manual_colour">
	       <repeat name="colours" title="Colours for the plots" min="1" max="50">
	       <param name="annotation_color" type="color" label="Colours" value="#ff00ff" help="Numbers of colours should be the same as number of components">
	       <sanitizer>
	           <valid initial="string.letters,string.digits">
	           <add value="#" />
	           </valid>
	       </sanitizer>
	       </param>
	       </repeat>
	    </when>
	    <when value="colourpalette">
	        <param name="palettes" type="select" display="radio" label="Select a colourpalette">
		    <option value="hue_pal()" selected="True">hue</option>
		    <option value="rainbow">rainbow</option>
		    <option value="heat.colors">heat colors</option>
		    <option value="terrain.colors">terrain colors</option>
		    <option value="topo.colors">topo colors</option>
		    <option value="cm.colors">cm colors</option>
	        </param>
	    </when>
        </conditional>
        <param name="output_rdata" type="boolean" label="Results as .RData output" help="Can be used to generate a classification prediction on new data"/>
    </inputs>
    <outputs>
        <data format="pdf" name="classification_images" from_work_dir="classificationpdf.pdf" label = "${tool.name} on ${on_string}: results"/>
        <data format="tabular" name="mzfeatures" label="${tool.name} on ${on_string}: features">
            <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'spatialShrunkenCentroids' and type_cond['method_cond']['ssc_analysis_cond']['ssc_method'] == 'ssc_analysis' or type_cond['type_method'] == 'prediction' and type_cond['classification_type_cond']['classification_type'] == 'SSC_classifier'</filter>
        </data>
        <data format="tabular" name="pixeloutput" label="${tool.name} on ${on_string}: pixels">
            <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'PLS' and type_cond['method_cond']['analysis_cond']['PLS_method'] == 'PLS_analysis' or type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'OPLS' and type_cond['method_cond']['opls_analysis_cond']['opls_method'] == 'opls_analysis' or type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'spatialShrunkenCentroids' and type_cond['method_cond']['ssc_analysis_cond']['ssc_method'] == 'ssc_analysis' or type_cond['type_method'] == 'prediction'</filter>
        </data>
        <data format="tabular" name="coefficients" label="${tool.name} on ${on_string}: coefficients">
            <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'PLS' and type_cond['method_cond']['analysis_cond']['PLS_method'] == 'PLS_analysis' or type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'OPLS' and type_cond['method_cond']['opls_analysis_cond']['opls_method'] == 'opls_analysis' or type_cond['type_method'] == 'prediction' and type_cond['classification_type_cond']['classification_type'] == 'PLS_classifier' or type_cond['type_method'] == 'prediction' and type_cond['classification_type_cond']['classification_type'] == 'OPLS_classifier'</filter>
        </data>
        <data format="tabular" name="loadings_weights" label="${tool.name} on ${on_string}: loadings and weights">
            <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'PLS' and type_cond['method_cond']['analysis_cond']['PLS_method'] == 'PLS_analysis' or type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'OPLS' and type_cond['method_cond']['opls_analysis_cond']['opls_method'] == 'opls_analysis' or type_cond['type_method'] == 'prediction' and type_cond['classification_type_cond']['classification_type'] == 'PLS_classifier' or type_cond['type_method'] == 'prediction' and type_cond['classification_type_cond']['classification_type'] == 'OPLS_classifier'</filter>
        </data>
        <data format="txt" name="best_r" label="${tool.name} on ${on_string}:best r">
        <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'spatialShrunkenCentroids' and type_cond['method_cond']['ssc_analysis_cond']['ssc_method'] == 'ssc_cvapply' and type_cond['method_cond']['ssc_analysis_cond']['write_best_params']</filter>
        </data>
        <data format="txt" name="best_s" label="${tool.name} on ${on_string}:best s">
        <filter>type_cond['type_method'] == 'training' and type_cond['method_cond']['class_method'] == 'spatialShrunkenCentroids' and type_cond['method_cond']['ssc_analysis_cond']['ssc_method'] == 'ssc_cvapply' and type_cond['method_cond']['ssc_analysis_cond']['write_best_params']</filter>
        </data>
        <data format="rdata" name="classification_rdata" label="${tool.name} on ${on_string}: results.RData">
        <filter>output_rdata</filter>
        </data>
    </outputs>
    <tests>
        <test expect_num_outputs="1">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "pixel_annotation_file1.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="4"/>
                <param name="column_fold" value="3"/>
                <param name="tabular_header" value="False"/>
                <conditional name="method_cond">
                    <param name="class_method" value="PLS"/>
                    <conditional name="analysis_cond">
                        <param name="PLS_method" value="cvapply"/>
                        <param name="plscv_comp" value="2:4"/>
                    </conditional>
                </conditional>
            </conditional>
            <output name="classification_images" file="test1.pdf" compare="sim_size" delta="2000"/>
        </test>

        <test expect_num_outputs="5">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "pixel_annotation_file1.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="4"/>
                <param name="tabular_header" value="False"/>
                <conditional name="method_cond">
                    <param name="class_method" value="PLS"/>
                    <conditional name="analysis_cond">
                        <param name="PLS_method" value="PLS_analysis"/>
                        <param name="pls_comp" value="2"/>
                        <param name="pls_scale" value="TRUE"/>
                        <param name="PLS_Yweights" value="TRUE"/>
                        <!--param name="pls_toplabels" value="100"/-->
                    </conditional>
                </conditional>
            </conditional>
            <param name="output_rdata" value="True"/>
            <output name="coefficients">
                <assert_contents>
                    <has_text text="900.004699707031"/>
                    <has_text text="962.870727539062"/>
                    <has_text text="999.606872558594"/>
                </assert_contents>
            </output>
            <output name="loadings_weights">
                <assert_contents>
                    <has_text text="900.076354980469"/>
                    <has_text text="950.495910644531"/>
                    <has_text text="989.024536132812"/>
                </assert_contents>
            </output>
            <output name="pixeloutput" file="pixels_test2.tabular"/>
            <output name="classification_images" file="test2.pdf" compare="sim_size"/>
            <output name="classification_rdata" file="test2.rdata" compare="sim_size"/>
        </test>

        <test expect_num_outputs="1">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "random_factors.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="4"/>
                <param name="column_fold" value="3"/>
                <param name="tabular_header" value="False"/>
                <conditional name="method_cond">
                    <param name="class_method" value="OPLS"/>
                    <conditional name="opls_analysis_cond">
                        <param name="opls_method" value="opls_cvapply"/>
                        <param name="opls_cvcomp" value="1:2"/>
                    </conditional>
                </conditional>
            </conditional>
            <output name="classification_images" file="test3.pdf" compare="sim_size"/>
        </test>

        <test expect_num_outputs="5">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "random_factors.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="4"/>
                <param name="tabular_header" value="False"/>
                <conditional name="method_cond">
                    <param name="class_method" value="OPLS"/>
                    <conditional name="opls_analysis_cond">
                        <param name="opls_method" value="opls_analysis"/>
                        <param name="opls_comp" value="3"/>
                        <param name="opls_scale" value="FALSE"/>
                        <param name="PLS_Yweights" value="FALSE"/>
                    </conditional>
                </conditional>
            </conditional>
            <param name="output_rdata" value="True"/>
            <output name="pixeloutput" file="pixels_test4.tabular"/>
            <output name="coefficients">
                <assert_contents>
                    <has_text text="900.148010253906"/>
                    <has_text text="974.132446289062"/>
                    <has_text text="999.908935546875"/>
                </assert_contents>
            </output>
            <output name="loadings_weights">
                <assert_contents>
                    <has_text text="901.581848144531"/>
                    <has_text text="939.189086914062"/>
                    <has_text text="984.185363769531"/>
                </assert_contents>
            </output>
            <output name="classification_images" file="test4.pdf" compare="sim_size"/>
            <output name="classification_rdata" file="test4.rdata" compare="sim_size"/>
        </test>

        <test expect_num_outputs="3">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "pixel_annotation_file1.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="3"/>
                <param name="column_fold" value="4"/>
                <param name="tabular_header" value="False"/>
                <conditional name="method_cond">
                    <param name="class_method" value="spatialShrunkenCentroids"/>
                    <conditional name="ssc_analysis_cond">
                        <param name="ssc_method" value="ssc_cvapply"/>
                        <param name="ssc_r" value="1:2"/>
                        <param name="ssc_s" value="2:3"/>
                        <param name="ssc_kernel_method" value="adaptive"/>
                        <param name="write_best_params" value="TRUE"/>
                    </conditional>
                </conditional>
            </conditional>
            <output name="classification_images" file="test5.pdf" compare="sim_size"/>
            <output name="best_r" file="best_r_test5.txt"/>
            <output name="best_s" file="best_s_test5.txt"/>
        </test>

        <test expect_num_outputs="4">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="training"/>
                <param name="annotation_file" value= "random_factors.tabular" ftype="tabular"/>
                <param name="column_x" value="1"/>
                <param name="column_y" value="2"/>
                <param name="column_response" value="4"/>
                <conditional name="method_cond">
                    <param name="class_method" value="spatialShrunkenCentroids"/>
                    <conditional name="ssc_analysis_cond">
                        <param name="ssc_method" value="ssc_analysis"/>
                        <param name="ssc_toplabels" value="20"/>
                     </conditional>
                    <param name="ssc_r" value="2"/>
                    <param name="ssc_s" value="2"/>
                    <param name="ssc_kernel_method" value="adaptive"/>
                </conditional>
            </conditional>
            <param name="output_rdata" value="True"/>
            <output name="mzfeatures" file="features_test6.tabular"/>
            <output name="pixeloutput" file="pixels_test6.tabular"/>
            <output name="classification_images" file="test6.pdf" compare="sim_size"/>
            <output name="classification_rdata" file="test6.rdata" compare="sim_size" delta="15000"/>
        </test>

        <test expect_num_outputs="5">
            <param name="infile" value="testfile_squares.rdata" ftype="rdata"/>
            <conditional name="type_cond">
                <param name="type_method" value="prediction"/>
                <param name="type_method" value="prediction"/>
                <param name="training_result" value="test2.rdata" ftype="rdata"/>
                <param name="classification_type" value="PLS_classifier"/>
                <conditional name="new_y_values_cond">
                    <param name="new_y_values" value="new_response"/>
                        <param name="new_response_file" value="pixel_annotation_file1.tabular" ftype="tabular"/>
                        <param name="column_new_x" value="1"/>
                        <param name="column_new_y" value="2"/>
                        <param name="column_new_response" value="4"/>
                        <param name="new_tabular_header" value="False"/>
                </conditional>
            </conditional>
            <param name="output_rdata" value="True"/>
            <output name="coefficients" file="coefficients_test7.tabular"/>
            <output name="loadings_weights" file="loadings_and_weights_test7.tabular"/>
            <output name="pixeloutput" file="pixels_test7.tabular"/>
            <output name="classification_images" file="test7.pdf" compare="sim_size"/>
            <output name="classification_rdata" file="test7.rdata" compare="sim_size" />
        </test>
    </tests>
    <help>
        <![CDATA[


@CARDINAL_DESCRIPTION@

-----

This tool provides three different Cardinal functions for supervised classification of mass-spectrometry imaging data.

@MSIDATA_INPUT_DESCRIPTION@
            - NA intensities are not allowed
            - duplicated coordinates will be removed

- For training: tabular file with condition and fold for each pixel: Two columns for pixel coordinates (x and y values); one column with the condition for the pixel, which will be used for classification; for the cross validation (cvapply) another column with a fold is necessary, each fold must contain pixels of all response groups and is used for cross validation. Condition and fold columns are treated as factor to perform discriminant analysis (also when numeric values are provided).

    ::

     x_coord     y_coord      condition    fold
        1            1           A          f1
        2            1           A          f2
        3            1           A          f3
        1            2           B          f1
        2            2           B          f2
        3            2           B          f3
       ...
       ...


- For prediction: RData output from previous classification run is needed as input, optionally new response values can be loaded with a tabular file containing x values, y values and the response


**Options**

- PLS-DA: partial least square discriminant analysis
- O-PLS-DA: Orthogonal partial least squares discriminant analysis
- Spatial shrunken centroids (more details in `Bemis et al. <https://doi.org/10.1074/mcp.O115.053918>`_)
- training and prediction

    - training can be done with cvapply that uses cross validation to find the best value for s, this requires not only a condition for each spectrum but also a fold (each fold should contain spectra of all conditions)
    - training with the best value for r and s gives the top m/z features for each condition and the predicted classification group for each spectrum
    - training result can be saved as RData file that can be reused for prediction of further samples
    - prediction can calculate accuracies when the annotations are known and provided


.. image:: $PATH_TO_IMAGES/classification_overview.png
   :width: 1000
   :height: 465



**Tips**

- The classification function will only run on files with valid intensity values (NA are not allowed)
- Only a single input file is accepted, several files have to be combined previously, for example with the MSI combine tool.


**Output**

- Pdf with the heatmaps and plots for the classification
- Tabular file with information on m/z features and pixels: toplabels/classes
- Optional: RData output that can be used to predict new data or to explore the results more deeply with the Cardinal package in R

        ]]>
    </help>
    <expand macro="citations"/>
</tool>