flupipe.Rmd

---
title: "Flupipe Report"
output:
  html_document:
    toc: true
    toc_float: true
    toc_depth: 6
    mathjax: null
date: "`r format(Sys.time(), '%d %B, %Y')`"
params:
  proj_folder: "none"
  list_folder: "none"
  min_cov: "none"
  reference: false
  rki_sample_name: true
  run_name: "none"
  version: "none"
---

<style>
  .superwideimage{
      overflow-x:scroll;
      white-space: nowrap;
  }

  .superwideimage img{
     max-width: none;
  }

</style>

<style>
  .superhighimage{
      overflow:auto;
      height: 500px;
      width: 100%;
      margin-top: 10px;
      margin-bottom:
  }

</style>

<style type="text/css">
.main-container {
  max-width: 100% !important;
  margin: auto;
}
</style>

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE, message = FALSE, warning = FALSE, error = FALSE, time_it = TRUE)

library("data.table")
library("formattable")
library("ggplot2")
library("kableExtra")
library("plyr")
library("rjson")
library("reshape2")
library("ShortRead")
library("dplyr")

min.coverage <- as.numeric(params$min_cov)
reference <- params$reference
stepsize = 20 # number of samples per plot 
n_content_threshold = 5 # threshold of N content % that should be marked in the table
```

```{r get_cmd_line_parameters}
# find all folders of intermediate results
l.dirlist <- list.dirs(file.path(params$proj_folder, "intermediate_data"), recursive = F)

# get the precise folder name of each interesting subfolder needed for tables
trimmed_folder <- l.dirlist[grepl(pattern = "trimmed", x = l.dirlist)]
mapping_stats_folder <- l.dirlist[grepl(pattern = "mapping_stats", x = l.dirlist)]
mapping_folder <- l.dirlist[grepl(pattern = "04_mapp*", x = l.dirlist)]
kraken_folder <- l.dirlist[grepl(pattern = "classified", x = l.dirlist)]

# get folder name for fasta files 
fasta.dir <- file.path(params$proj_folder, "consensuses_iupac")

if (!dir.exists(params$proj_folder)) {
    stop("input folder does not exist")
}

if (!dir.exists(params$list_folder)) {
    stop("output folder does not exist")
}

if (!dir.exists(trimmed_folder)) {
    stop("trimmed folder does not exist")
}
if (!dir.exists(mapping_stats_folder)) {
    stop("mapping stats folder does not exist")
}

if (!dir.exists(mapping_folder)) {
    stop("mapping folder does not exist")
}

if (!dir.exists(fasta.dir)) {
    stop("consensus folder with fasta files does not exist")
}

kraken_run <- FALSE
if (!dir.exists(kraken_folder) || length(kraken_folder) == 0) {
    kraken_run <- FALSE
} else {
    kraken_run <- TRUE
}
```

```{r def_functions}
# input of file listing. requires file names like /path/to/nCoV1.bamstats.txt
f.get_filenames <- function(l.input){
    sapply(l.input, function(x){
        filename <- unlist(strsplit(x, '/'))[length(unlist(strsplit(x, '/')))]
        samplename <- unlist(strsplit(filename, '\\.'))[1]
        return(samplename)
    })
    #returns named character vector of sample names, e.g. nCoV1
}

# modified color_bar to fix direction from "rtl" to "ltr"
f.color_bar <- function (color = "lightgray", fun = "proportion",
    ...)
{
    fun <- match.fun(fun)
    formattable::formatter("span", style = function(x) style(display = "inline-block",
        direction = "ltr", `border-radius` = "4px",
        `padding-right` = "2px", `background-color` = csscolor(color),
        width = percent(fun(as.numeric(x), ...))))
}

  # Function to collapse the values within a grouped dataframe
collapse_rows_df <- function(df, variable){
    group_var <- enquo(variable)
    df %>%
      group_by(!! group_var) %>%
      mutate(groupRow = 1:n()) %>%
      ungroup() %>%
      mutate(!!quo_name(group_var) := ifelse(groupRow == 1, 
      as.character(!! group_var), "")) %>%
      select(-c(groupRow))
  }
```

```{r read_data}
# this chunk reads mapping stats and trimming stats
l.infiles.trimming <- list.files(path = trimmed_folder,
                                 pattern = "*.fastp.json$",
                                 full.names = T,
                                 recursive = T)
l.infiles.bamstats <- list.files(path = mapping_stats_folder,
                                 pattern = "*.bamstats.pipe.txt$",
                                 full.names = T,
                                 recursive = T)

l.infiles.coverage <- list.files(path = mapping_stats_folder,
                                 pattern = "*.coverage.tsv$",
                                 full.names = T,
                                 recursive = T)

l.infiles.fragsize <- list.files(path = mapping_stats_folder,
                                 pattern = "*.fragsize.tsv$",
                                 full.names = T,
                                 recursive = T)

l.infiles.bamstats.coverage <- list.files(path = mapping_stats_folder,
                                 pattern = "*.bamstats_coverage.txt$",
                                 full.names = T,
                                 recursive = T)

l.infiles.read.duplicates <- list.files(path = mapping_folder,
                                 pattern = "*.dedup.txt$",
                                 full.names = T,
                                 recursive = T)

l.infiles.read.ranking.refs <- list.files(path = mapping_folder,
                                 pattern = glob2rx("Ranking_top5_Refs_*.txt$"),
                                 full.names = T,
                                 recursive = T)
l.infiles.fasta.files <- list.files(path = fasta.dir,
                                 pattern = glob2rx("*.fasta$"),
                                 full.names = T,
                                 recursive = T)

l.infiles.merged.read.length <- list.files(path = trimmed_folder,
                                 pattern = "merged_length.tsv$",
                                 full.names = T,
                                 recursive = T)

# initialise empty list
l.infiles.kraken <- list()
if (kraken_run) {
    l.infiles.kraken <- list.files(path = kraken_folder,
                                   pattern = "*.report.txt$",
                                   full.names = T,
                                   recursive = T)
}
# put names to file names

names(l.infiles.trimming) <- f.get_filenames(l.infiles.trimming)
names(l.infiles.bamstats) <- f.get_filenames(l.infiles.bamstats)
names(l.infiles.coverage) <- f.get_filenames(l.infiles.coverage)
names(l.infiles.fragsize) <- f.get_filenames(l.infiles.fragsize)
names(l.infiles.kraken) <- f.get_filenames(l.infiles.kraken)
names(l.infiles.bamstats.coverage) <- f.get_filenames(l.infiles.bamstats.coverage)
names(l.infiles.read.duplicates) <- f.get_filenames(l.infiles.read.duplicates)
names(l.infiles.read.ranking.refs) <- f.get_filenames(l.infiles.read.ranking.refs)
names(l.infiles.fasta.files) <- f.get_filenames(l.infiles.fasta.files)
names(l.infiles.merged.read.length) <- f.get_filenames(l.infiles.merged.read.length)

if (length(l.infiles.merged.read.length) == 0) {
    stop("no merged read length file identified")
}

if (length(l.infiles.trimming) == 0) {
    stop("no trimming stat files identified")
}

if (length(l.infiles.bamstats) == 0) {
    stop("no bam stat files identified")
}

if (length(l.infiles.coverage) == 0) {
    stop("no coverage files identified")
}

if (length(l.infiles.fragsize) == 0) {
    stop("no fragment size files identified")
}

if (length(l.infiles.bamstats.coverage) == 0) {
    stop("no bamstats coverage size files identified")
}

if (length(l.infiles.read.duplicates) == 0) {
    stop("no read duplicates size files identified")
}

if (length(l.infiles.read.ranking.refs) == 0) {
    stop("no Ranking_top5_Refs files identified")
}

if (length(l.infiles.fasta.files) == 0) {
    stop("no iupac consensus fasta files identified")
}

if (length(l.infiles.kraken) == 0) {
    kraken_run <- FALSE
}
```

## Sequencing run

```{r conditional_sentence, results='asis'}
if (params$run_name != 'none') {
  cat(paste0("Sequencing was performed in run: ", params$run_name, "\n"))
} else {
  cat(paste0("No ID for a sequencing run was submitted. \n"))
}
```

```{r read_coverage_preanalyses}
# read file content
# dt.coverage is used again later for plotting
dt.coverage <- as.data.table(ldply(l.infiles.coverage, fread, sep='\t'))
colnames(dt.coverage) <- c("sample","chromosome", "position", "depth")

# rename chromosome column values
dt.coverage$chromosome <- sapply(dt.coverage$chromosome, function(x){
  segment <- tail(unlist(strsplit(x,"|",fixed=TRUE)), n=1)
  segment <- toString(segment)
  return(segment)
})

dt.output <- dt.coverage[, sum(depth > 10), by = sample]
setnames(dt.output, "V1", "covered.bases")
dt.output$genome.length <- dt.coverage[,length(depth), by = sample]$V1
dt.output$genome.coverage <- dt.output$covered.bases / dt.output$genome.length

### taking directly the per chromosome results from bamstats coverage
### this data frame is used again later for the whole table including all samples

dt.bamstats.coverage <- as.data.table(ldply(l.infiles.bamstats.coverage, fread, sep='\t', select=c(1,4,6,7)))
dt.bamstats.coverage[,("coverage") := round(.SD,2), .SDcols="coverage"]
dt.bamstats.coverage[,("meandepth") := round(.SD,0), .SDcols="meandepth"]
colnames(dt.bamstats.coverage) <- c("sample","chromosome","reads mapped [%]", "coverage [%]", "mean depth [bp]")

# rename chromosome column values
dt.bamstats.coverage$chromosome <- sapply(dt.bamstats.coverage$chromosome, function(x){
  segment <- tail(unlist(strsplit(x,"|",fixed=TRUE)), n=1)
  segment <- toString(segment)
  return(segment)
})

dt.bamstats.coverage <- reshape(dt.bamstats.coverage, idvar = "sample", timevar = "chromosome", direction = "wide")

# adding placeholder column(s) for missing segment(s)
ls.segment <- list("*HA","*NA","*MP","*NP","*NS","*PA","*PB1","*PB2")
substrRight <- function(x, n){substr(x, nchar(x)-n+1, nchar(x))}

if (ncol(dt.bamstats.coverage) != 25 ) {
    for (s in ls.segment){
      if (Reduce("|",grepl(s, names(dt.bamstats.coverage)))) {}
      else {
        dt.bamstats.coverage[,paste("reads mapped [%].", substrRight(s, 2))] <- NA
        dt.bamstats.coverage[,paste("coverage [%].", substrRight(s, 2))] <- NA
        dt.bamstats.coverage[,paste("mean depth [bp].", substrRight(s, 2))] <- NA
        }
  }
}

# reorder dataframe columns
dt.bamstats.coverage <- dt.bamstats.coverage[, c( "sample","reads mapped [%].HA","coverage [%].HA","mean depth [bp].HA",
                                                  "reads mapped [%].NA","coverage [%].NA","mean depth [bp].NA",
                                                  "reads mapped [%].MP","coverage [%].MP","mean depth [bp].MP",
                                                  "reads mapped [%].NP","coverage [%].NP","mean depth [bp].NP",
                                                  "reads mapped [%].NS","coverage [%].NS","mean depth [bp].NS",
                                                  "reads mapped [%].PA","coverage [%].PA","mean depth [bp].PA",
                                                  "reads mapped [%].PB1","coverage [%].PB1","mean depth [bp].PB1",
                                                  "reads mapped [%].PB2","coverage [%].PB2","mean depth [bp].PB2" )]

###################### Joining the 2 datasets ##################################
colnames(dt.output) <- c("sample", "ref.coverage [bp]", "genome.length", "ref.coverage [fraction]")
dt.output <- merge(dt.output, dt.bamstats.coverage, by = "sample")
```

```{r conditional_warning, results='asis'}
dt.output <- dt.output[grepl("NK|Empty", dt.output$sample, ignore.case = TRUE) & dt.output$`ref.coverage [bp]` >= 0.2, ]
if (nrow(dt.output) > 0) {

    cat("## WARNING \n")
    cat("Samples identified automatically as negative controls show unusual large coverage of the reference genome (more than 20%). Please check the following table, if this is expected for the samples. \n")
}
```


```{r conditional_table}
if (nrow(dt.output) > 0) {
    dt.output$`ref.coverage [fraction]` <- round(dt.output$`ref.coverage [fraction]`, 2)
    # add a row with all 0 and 1 to make colour scaling reproducible
    df.tmp <- data.frame("sample" = c(0,1),
                         "genome.length" = c(0,1),
                         "ref.coverage [bp]" = c(0,1),
                         "ref.coverage [fraction]" = c(0,1),
                         "reads mapped [%].HA" = c(0,1),
                         "coverage [%].HA" = c(0,1),
                         "mean depth [bp].HA" = c(0,1), 
                         "reads mapped [%].NA" = c(0,1),
                         "coverage [%].NA" = c(0,1),
                         "mean depth [bp].NA" = c(0,1),
                         "reads mapped [%].MP" = c(0,1),
                         "coverage [%].MP" = c(0,1),
                         "mean depth [bp].MP" = c(0,1),
                         "reads mapped [%].NP" = c(0,1),
                         "coverage [%].NP" = c(0,1),
                         "mean depth [bp].NP" = c(0,1),
                         "reads mapped [%].NS" = c(0,1),
                         "coverage [%].NS" = c(0,1),
                         "mean depth [bp].NS" = c(0,1),
                         "reads mapped [%].PA" = c(0,1),
                         "coverage [%].PA" = c(0,1),
                         "mean depth [bp].PA" = c(0,1),
                         "reads mapped [%].PB1" = c(0,1),
                         "coverage [%].PB1" = c(0,1),
                         "mean depth [bp].PB1" = c(0,1),
                         "reads mapped [%].PB2" = c(0,1),
                         "coverage [%].PB2" = c(0,1),
                         "mean depth [bp].PB2" = c(0,1)
                         )
    dt.output <- rbind(dt.output, df.tmp, use.names=FALSE)
    
    rm(df.tmp)
    # highlight important rows
    dt.output$sample <- cell_spec(dt.output$sample, color = ifelse(dt.output$`ref.coverage [fraction]` > 0.2, "red", "black"))
    dt.output$`ref.coverage [fraction]` <- color_tile("white", "orange")(dt.output$`ref.coverage [fraction]`)
    #remove added lines
    dt.output <- head(dt.output, n = -2)
    dt.output <- dt.output[,-c("genome.length")]

    kbl(dt.output,
        digits = 2,
        caption = "Reference genome coverage in base pairs and as fraction of the total length. Bases with more than 10x sequencing depth are counted as covered. Negative control's names are highlighted in red if more than 20% of reference genome is covered, increasing coverage fraction is coloured orange (scaled from 0 to 1).",
        escape = F) %>%
        kable_styling(bootstrap_options = c("striped", "hover"), fixed_thead = T) %>%
        scroll_box(height = "400px")
}
```

```{r conditional_cleanup}
rm(dt.output)
```

## Trimming statistics

Quality trimming and adapter clipping was performed using fastp (v 0.20.0).

### Read counts

Raw reads were subjected to adapter clipping.
The following table summarizes the read count per sample before and after trimming. Additionally, the percentage of remaining reads after trimming is listed.

```{r read_trimming}
l.trimming.data.json <- lapply(l.infiles.trimming, function(x){
    fromJSON(file = x)
})

df.trimming.data <- ldply(l.trimming.data.json, function(e){
    df.before <- as.data.frame(do.call(rbind, e$summary$before_filtering))
    colnames(df.before) <- c("before.trimming")
    df.after  <- as.data.frame(do.call(rbind, e$summary$after_filtering))
    colnames(df.after) <- c("after.trimming")

    df.output <- data.frame(feature = rownames(df.before),
                            before = df.before$before.trimming,
                            after = df.after$after.trimming)
    return(df.output)
})

# rename 1st column
tmp <- colnames(df.trimming.data)
tmp[1] <- c("sample")
colnames(df.trimming.data) <- tmp

df.filter.data <- ldply(l.trimming.data.json, function(e){

    df.output <- data.frame(passed_filter = e$filtering_result$passed_filter_reads,
                            low_qual = e$filtering_result$low_quality_reads,
                            high_N = e$filtering_result$too_many_N_reads,
                            low_complex = e$filtering_result$low_complexity_reads,
                            short = e$filtering_result$too_short_reads
                            )
    return(df.output)
})

# rename 1st column
tmp <- colnames(df.filter.data)
tmp[1] <- c("sample")
colnames(df.filter.data) <- tmp
```


```{r table_trimming}
df.summary <- data.frame(sample = unique(df.trimming.data$sample),
                        reads.before.clip = df.trimming.data$before[grepl("total_reads", df.trimming.data$feature)],
                        reads.after.clip  = df.trimming.data$after[grepl("total_reads", df.trimming.data$feature)],
                        percentage.after.clip = df.trimming.data$after[grepl("total_reads", df.trimming.data$feature)]*100/df.trimming.data$before[grepl("total_reads", df.trimming.data$feature)]
)
df.table <- df.summary
# add coloured bar charts to table
df.table$reads.before.clip <- f.color_bar("lightgreen")(df.table$reads.before.clip)
df.table$reads.after.clip <- f.color_bar("lightgreen")(df.table$reads.after.clip)
df.table$percentage.after.clip <- Map(paste,format(round(df.table$percentage.after.clip,2),nsmall=2),"%")

kbl(x = df.table,
    col.names = c("sample", "reads before trim", "reads after trim", "% of reads after trimming"),
    digits = 2,
    escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), fixed_thead = T, full_width = T) %>%
  scroll_box(height = "400px")

# save table as csv for later use
write.csv(  x=df.summary,
            row.names = FALSE,
            file = file.path(params$list_folder, "read_stats.csv")
          )
```

<div class="superhighimage">

```{r count_trimming_samples}
sample.count <- length(unique(df.table$sample))
plot.width <- 20
```

For a better overview, the reads counts are plotted in a bar plot per sample. The number of raw reads is shown in light blue and dark blue visualizes trimmed reads.

```{r plot_trimming}
if (sample.count > stepsize) {
  # for more than 20 samples split plot into multiple plots with each only 20 samples  
  for (s in 0:(ceiling(sample.count/stepsize)-1) ) {
    x = s*stepsize+1
    y = s*stepsize+stepsize
    df.summary.subset <- na.omit(df.summary[x:y,])
    df.plot <- melt(df.summary.subset, id.vars = "sample",
                    measure.vars = c("reads.before.clip", "reads.after.clip"))
    
    g <- ggplot(data = df.plot, aes(sample, value, fill = variable)) +
      geom_bar(stat = "identity", position = "dodge", width = 0.6) +
      labs(title = "read counts before & after trimming",
           x = "sample", y = "count", fill = "trimming") +
      scale_fill_manual(values = c("#56B4E9", "#0072B2"), labels = c("raw", "trimmed")) +
      scale_y_continuous(labels = function(x) format(x, scientific = F)) +
      theme(axis.text.x = element_text(angle = 90, hjust = 1),
            legend.position = "top") +
      coord_flip()
    print(g)
  }
} else {
    df.plot <- melt(df.summary, id.vars = "sample",
                    measure.vars = c("reads.before.clip", "reads.after.clip"))
    ggplot(data = df.plot, aes(sample, value, fill = variable)) +
        geom_bar(stat = "identity", position = "dodge", width = 0.6) +
        labs(title = "read counts before & after trimming",
             x = "sample", y = "count", fill = "trimming") +
        scale_fill_manual(values = c("#56B4E9", "#0072B2"), labels = c("raw", "trimmed")) +
        scale_y_continuous(labels = function(x) format(x, scientific = F)) +
        theme(axis.text.x = element_text(angle = 90, hjust = 1),
              legend.position = "top") +
        coord_flip()
}
```

</div>

### Read Length Distribution
For the forward (R1) and reverse (R2) read the sequence length distribution is shown after trimming.

``` {r read_len_table}
header <- read.table(l.infiles.merged.read.length, nrow = 1, header = TRUE)
column_name <- substr(colnames(header),2, nchar(colnames(header)))
setnames(header, column_name)

####### rearranging columns so that R1 and R2 of the same sample are next to each other #######
rearrenged_colnames <- c()
for (h in 1:(length(colnames(header))/2)) {
  a = length(colnames(header))/2+h
  rearrenged_colnames <- append(rearrenged_colnames, colnames(header)[h])
  rearrenged_colnames <- append(rearrenged_colnames, colnames(header)[a])
}
setcolorder(header, rearrenged_colnames)

reads <- as.data.table(ldply(l.infiles.merged.read.length, fread, sep='\t', skip=1, header=FALSE))
reads <- reads[,-1]
rearrenged_colnames_reads <- c()
for (h in 1:(length(colnames(reads))/2)) {
  a = length(colnames(reads))/2+h
  rearrenged_colnames_reads <- append(rearrenged_colnames_reads, colnames(reads)[h])
  rearrenged_colnames_reads <- append(rearrenged_colnames_reads, colnames(reads)[a])
}
setcolorder(reads, rearrenged_colnames_reads)
setnames(reads, colnames(header))
```

``` {r reduce_data_read_len}
# reducing the dataframe by sampling every 100th read/row
reads = reads[seq(1, nrow(reads), 50), ]
```
<div class=superhighimage>
``` {r reshape_and_plot_read_len}
ss = stepsize*2 # since we have 2 columns per sample (R1 and R2)
min_val <- as.numeric(min(na.omit(reads))-50)
if (min_val<0){min_val <- as.numeric(0)}
max_val <- as.numeric(max(na.omit(reads))+50)

if (length(colnames(reads)) > ss) {
  # for more than 20 samples split plot into multiple plots with each only 20 samples
  for (s in 0:(ceiling(length(colnames(reads))/ss)-1) ) {
    x = s*ss+1
    y = s*ss+ss
    if (y > length(colnames(reads))) {
      y = length(colnames(reads))
    }
    
    ####### Reshape dataframe for boxplot ########
    reads.subset <- reads[,x:y]
    reads.subset <- stack(as.data.frame(reads.subset))
    reads.subset$ind <- as.character(reads.subset$ind)
    
    reads.subset$sample <- substr(reads.subset$ind, 1, nchar(reads.subset$ind)-3)
    reads.subset$read <- substr(reads.subset$ind, nchar(reads.subset$ind)-1, nchar(reads.subset$ind))
    reads.subset <- reads.subset[,-2] # remove ind column
    # print(reads.subset)
    
    ####### Plot each 20 samples #######
    p <- ggplot(reads.subset) +
          geom_boxplot(aes(x = values, y = sample, fill = read),
                       outlier.shape = NA, na.rm=TRUE) +
          labs(title = "Sequence Length Distribution", x = "Sequence Legth (bp)",
               y = "Sample") +
          theme(axis.text.x = element_text(angle = 90, hjust = 1),
                legend.position = "top") +
          scale_fill_manual(values=c("lightgreen", "green4")) +
          xlim(min_val, max_val)
    print(p)
    rm(reads.subset)
  }
} else {
    read_df <- stack(as.data.frame(reads))
    read_df$ind <- as.character(read_df$ind)
    read_df$sample <- substr(read_df$ind, 1, nchar(read_df$ind)-3)
    read_df$read <- substr(read_df$ind, nchar(read_df$ind)-1, nchar(read_df$ind))
  
    ggplot(read_df) +
      geom_boxplot(aes(x = values, y = sample, fill = read), outlier.shape = NA, na.rm=TRUE) +
      labs(title = "Sequence Length Distribution", x = "Sequence Legth [bp]", y = "Sample") +
      theme(axis.text.x = element_text(angle = 90, hjust = 1), legend.position = "top")+
      scale_fill_manual(values=c("lightgreen", "green4")) +
      xlim(min_val, max_val)
}
```
</div>

### Read Quality Distribution
For the forward (R1) and reverse (R2) read the sequence quality distribution is shown after trimming. 

```{r read_trimming_for_phred}
df.trimming.phred.data <- ldply(l.trimming.data.json, function(e){
    df.output <- data.frame(phred_score_R1 = c(e$"read1_after_filtering"$"quality_curves"$"mean"),
                            phred_score_R2 = c(e$"read2_after_filtering"$"quality_curves"$"mean"))
    return(df.output)
})
colnames(df.trimming.phred.data) <- c("Sample", "R1", "R2")
```

<div class=superhighimage>
``` {r plot_phred_score}
samples <- unique(df.trimming.phred.data$Sample)
if (length(samples)>stepsize) {
  for (s in 0:(ceiling(length(samples)/stepsize)-1)){
    x= s*stepsize
    y = ((s+1)*stepsize)-1
    if (y>length(samples)){
      y = length(samples)
    }
    df.subset <- subset(df.trimming.phred.data, Sample %in% samples[x:y])
    df.subset <- melt(df.subset, id.vars = "Sample", measure.vars = c("R1", "R2"))

    b <- ggplot(df.subset) +
    geom_boxplot(aes(x = value, y = Sample, fill = variable), outlier.shape = NA, na.rm=TRUE, width = 0.4) +
    labs(title = "Per Sequence Quality Scores", x = "Mean Sequence Quality (Phred Score)", y = "Sample", fill = "read") +
    theme(axis.text.x = element_text(angle = 90, hjust = 1), legend.position = "top") +
    scale_fill_manual(values=c("lightgreen", "green4"))
    print(b)
    rm(df.subset)
  }
} else {
  
 df.trimming.phred.data <- melt(df.trimming.phred.data, id.vars = "Sample", measure.vars = c("R1", "R2"))
  ggplot(df.trimming.phred.data) +
    geom_boxplot(aes(x = value, y = Sample, fill = variable), outlier.shape = NA, na.rm=TRUE, width = 0.4) +
    labs(title = "Per Sequence Quality Scores", x = "Mean Sequence Quality (Phred Score)", y = "Sample", fill = "read") +
    theme(axis.text.x = element_text(angle = 90, hjust = 1), legend.position = "top") +
    scale_fill_manual(values=c("lightgreen", "green4"))
}
```

</div>

```{r trimming_cleanup}
rm(df.plot)
rm(df.summary)
rm(df.table)
rm(df.trimming.data)
rm(df.filter.data)
rm(reads)
rm(read_df)
rm(df.trimming.phred.data)
```

### Taxonomic Read Classification

The taxonomic classification of reads can not only serve to identify contamination, but also enable the filtering of reads assigned to certain taxa. In this step influena A, influenza B and Orthomyxoviridae reads were filtered and used for the next analysis step. Remaining reads were excluded from further analysis.

The assigned taxa are listed by sample n the following table.

```{r read_kraken}
    # Kraken2 output column labels.
    # Percentage of fragments covered by the clade rooted at this taxon
    # Number of fragments covered by the clade rooted at this taxon
    # Number of fragments assigned directly to this taxon
    # A rank code, indicating (U)nclassified, (R)oot, (D)omain, (K)ingdom, (P)hylum, (C)lass, (O)rder, (F)amily, (G)enus, or (S)pecies. Taxa that are not at any of these 10 ranks have a rank code that is formed by using the rank code of the closest ancestor rank with a number indicating the distance from that rank. E.g., "G2" is a rank code indicating a taxon is between genus and species and the grandparent taxon is at the genus rank.
    # NCBI taxonomic ID number
    # Indented scientific name

df.kraken_output <- data.frame()
if (kraken_run) {
    dt.kraken_data <- ldply(l.infiles.kraken, fread)
    colnames(dt.kraken_data) <- c("Sample", "read_ratio", "read_count", "read_count_specific", "rank", "ncbi_taxid", "sciname")
    # select unclassified and tax id's for Orthomyxoviridae (11308), Influenza A (11320) and B (11520), and Human (9606)
    df.kraken_output <- dt.kraken_data[dt.kraken_data$ncbi_taxid %in% c(0,"11308","11320","11520","9606"), c("Sample", "read_ratio", "read_count", "ncbi_taxid", "sciname")]

    # make tables wide for absolute and relative read counts
    dt.kraken.count <- data.table::dcast(as.data.table(df.kraken_output), Sample ~ sciname, value.var = c("read_count"))
    dt.kraken.count[is.na(dt.kraken.count)] <- 0
    df.kraken_output <- data.table::dcast(as.data.table(df.kraken_output), Sample ~ sciname, value.var = c("read_ratio"))
    df.kraken_output[is.na(df.kraken_output)] <- 0
    
    # Add placeholder for missing columns 
    species <- c("Sample","Homo sapiens", "Influenza A virus", "Influenza B virus", "Orthomyxoviridae", "unclassified")
    #df.kraken_output
    for (sp in species) {
      if (Reduce("|",grepl(sp, colnames(df.kraken_output)))) {
        # do nothing if column exist
      }
      else {
        df.kraken_output <- cbind(df.kraken_output, new_col = 0)
        colnames(df.kraken_output)[which(names(df.kraken_output) == "new_col")] <- sp
        }
    }
    #dt.kraken.count
     for (sp in species) {
      if (Reduce("|",grepl(sp, colnames(dt.kraken.count)))) {
        # do nothing if column exist
      }
      else {
        dt.kraken.count <- cbind(dt.kraken.count, new_col = 0)
        colnames(dt.kraken.count)[which(names(dt.kraken.count) == "new_col")] <- sp
        }
    }  
    colnames(df.kraken_output) <- paste0(colnames(df.kraken_output), c("", " [%]", " [%]", " [%]", " [%]", " [%]" ))
    
    df.kraken_output$`filter passed` <- rowSums(dt.kraken.count[,c("Homo sapiens", "Influenza A virus", "Influenza B virus")]) *2
    #print(colnames(df.kraken_output))
    df.kraken_output <- df.kraken_output[, c("Sample", "filter passed", "Homo sapiens [%]", "Influenza A virus [%]", "Influenza B virus [%]", "Orthomyxoviridae [%]", "unclassified [%]")]
}
```

```{r table_kraken, fig.cap="Read counts after species binning using Kraken."}
if (kraken_run) {
      # save table as csv for later use
    write.csv(  x=df.kraken_output,
                row.names = FALSE,
                file = file.path(params$list_folder, "species_filtering.csv")
    )

    kbl(df.kraken_output) %>%
        kable_styling(bootstrap_options = c("striped", "hover"), fixed_thead = T) %>%
        scroll_box(height = "400px")
}
```

```{r reshape_kraken_barchart}
if (kraken_run) {
    # reshape dataframe for  plotting
        # add column for diff of ortho and infA + infB (for barplot)
    df.kraken_output$diff_orthomyxoviridae  <- df.kraken_output$`Orthomyxoviridae [%]`-(df.kraken_output$`Influenza A virus [%]` + df.kraken_output$`Influenza B virus [%]`)
    
    df.kraken_output.reshaped <- reshape(df.kraken_output,
                     varying=names(df.kraken_output)[3:8],
                     direction="long", idvar=c("Sample"),
                     v.names="ratio", timevar="tax")
    df.kraken_output.reshaped$tax <- factor(df.kraken_output.reshaped$tax,
                                      levels = c(1, 2, 3, 4, 5, 6),
                                      labels = names(df.kraken_output)[3:8])
    df.kraken_output.reshaped$Sample <- factor(df.kraken_output.reshaped$Sample)
    df.kraken_output.reshaped$tax <- factor(df.kraken_output.reshaped$tax)
}
```



<div class=superhighimage>
```{r kraken_plot_barchart, fig.align='center', fig.cap="Taxonimical Classification of the Samples."}
if (kraken_run) {
  df.kraken.barplot <- df.kraken_output.reshaped[df.kraken_output.reshaped$tax %in% c("Homo sapiens [%]", "Influenza A virus [%]", "Influenza B virus [%]", "unclassified [%]", "diff_orthomyxoviridae"), ]
  
  samples <- unique(df.kraken.barplot$Sample)
  if (length(samples)>stepsize) {
    for (s in 0:(ceiling(length(samples)/stepsize)-1)){
      x = s*stepsize
      y = ((s+1)*stepsize)-1
      if (y>length(samples)){y = length(samples)}
      df.subset <- subset(df.kraken.barplot, Sample %in% samples[x:y])
      
      g <- ggplot(data=df.subset, aes(fill=tax, y=Sample, x=ratio)) +
          geom_bar(position="stack", stat="identity", width = 0.6) +
          labs(title = "Taxanomic Classification", fill = "") +
          theme(legend.position = 'top', legend.text = element_text(size = 7),
                legend.title = element_text(size=8)) +
          guides(fill = guide_legend(label.position = "bottom",
                                     title.position = "left", title.vjust = 1)) +
          scale_fill_brewer(palette="Set1", labels = c("Homo Sapiens", "Influenza A", "Influenza B", "unclassified", "Orthomyxoviridae"))
      print(g)
    }
  } else {
    ggplot(data=df.kraken.barplot, aes(fill=tax, y=Sample, x=ratio)) +
          geom_bar(position="stack", stat="identity", width = 0.6) +
          labs(title = "Taxanomic Classification", fill = "") +
          theme(legend.position = 'top', legend.text = element_text(size = 7),
                legend.title = element_text(size=8)) +
          guides(fill = guide_legend(label.position = "bottom",
                                     title.position = "left", title.vjust = 1)) +
          scale_fill_brewer(palette="Set1", labels = c("Homo Sapiens", "Influenza A", "Influenza B", "unclassified", "Orthomyxoviridae"))
  }
}
```
</div>

```{r kraken_cleanup}
if (kraken_run) {
  rm(dt.kraken.count)
  rm(df.kraken_output)
  rm(dt.kraken_data)
  rm(df.kraken_output.reshaped)
  rm(df.kraken.barplot)
}
```

## Mapping statisics

Reads were mapped to the reference genome using BWA. For each sample the number of mapped read as well as the overall mapping rate and the rate of duplicated reads are listed in the following table. Additionally, the genome coverage, the mean read depth and the rate of mapped reads are listed for each genome fragment.

```{r read_mappingstats}
df.bamstat.data <- ldply(l.infiles.bamstats, fread, sep = '|')
colnames(df.bamstat.data) <- c("sample", "count", "unknown", "description")
```

```{r read_mappingstats_bamstats_coverage}
# # read file content
# dt.bamstats.coverage <- as.data.table(ldply(l.infiles.bamstats.coverage, fread, sep='\t', select=c(1,4,6,7)))
# dt.bamstats.coverage[,("coverage") := round(.SD,2), .SDcols="coverage"]
# dt.bamstats.coverage[,("meandepth") := round(.SD,0), .SDcols="meandepth"]
# colnames(dt.bamstats.coverage) <- c("sample","chromosome","reads mapped [%]", "coverage [%]", "mean depth [bp]")
# 
# cc <- unlist(strsplit(dt.bamstats.coverage$chromosome,"|",fixed=TRUE))
# dt.bamstats.coverage$chromosome <- cc[4*(1:length(dt.bamstats.coverage$chromosome))]
# 
# dt.bamstats.coverage <- reshape(dt.bamstats.coverage, idvar = "sample", timevar = "chromosome", direction = "wide")
# 
# # adding placeholder column(s) for missing segment(s)
# ls.segment <- list("*HA","*NA","*MP","*NP","*NS","*PA","*PB1","*PB2")
# substrRight <- function(x, n){substr(x, nchar(x)-n+1, nchar(x))}
# 
# if (ncol(dt.bamstats.coverage) != 25 ) {
#     for (s in ls.segment){
#       if (Reduce("|",grepl(s, names(dt.bamstats.coverage)))) {}
#       else {
#         dt.bamstats.coverage[,paste("reads mapped [%].", substrRight(s, 2))] <- NA
#         dt.bamstats.coverage[,paste("coverage [%].", substrRight(s, 2))] <- NA
#         dt.bamstats.coverage[,paste("mean depth [bp].", substrRight(s, 2))] <- NA
#         }
#   }
# }
# 
# # reorder dataframe columns
# dt.bamstats.coverage <- dt.bamstats.coverage[, c( "sample","reads mapped [%].HA","coverage [%].HA","mean depth [bp].HA",
#                                                   "reads mapped [%].NA","coverage [%].NA","mean depth [bp].NA",
#                                                   "reads mapped [%].MP","coverage [%].MP","mean depth [bp].MP",
#                                                   "reads mapped [%].NP","coverage [%].NP","mean depth [bp].NP",
#                                                   "reads mapped [%].NS","coverage [%].NS","mean depth [bp].NS",
#                                                   "reads mapped [%].PA","coverage [%].PA","mean depth [bp].PA",
#                                                   "reads mapped [%].PB1","coverage [%].PB1","mean depth [bp].PB1",
#                                                   "reads mapped [%].PB2","coverage [%].PB2","mean depth [bp].PB2" )]

```

```{r read_duplicates}
# read file content
dt.read.duplicates <- as.data.table(ldply(l.infiles.read.duplicates, fread, sep='\t', select = c(9), skip=6, nrow=1)) 
colnames(dt.read.duplicates) <- c("sample","reads duplicated [%]")
dt.read.duplicates$`reads duplicated [%]` <- round(dt.read.duplicates$`reads duplicated [%]` * 100, digits = 2)
```

```{r table_bamstats}
df.output <- data.frame("sample" = unique(df.bamstat.data$sample),
                        "input" = df.bamstat.data$count[grepl("in total", df.bamstat.data$description)],
                        "mapped" = df.bamstat.data$count[grepl("properly paired", df.bamstat.data$description)])

df.output$mapping.rate <- round((df.output$mapped / df.output$input)*100, digits = 2)
colnames(df.output) <- c("sample", "reads in", "reads mapped", "mapping rate [%]")

# save table as csv for later use
write.csv(  x=df.output,
            row.names = FALSE,
            file = file.path(params$list_folder, "mapping_stats.csv")
)

# merge read per segment, duplication data with bamstats table
df.mapping.table <- merge(df.output, dt.read.duplicates, by="sample")
##### bamstats coverage table from above is reused 
df.mapping.table <- merge(df.mapping.table, dt.bamstats.coverage, by="sample")

# replace dots in column names with space
sub.dots <- function(df) {names(df) <- sub("\\.", " ", names(df));df}
df.mapping.table <- sub.dots(df.mapping.table)

# convert reads mapped per segment column values to mapping rate
seg_col_range <- seq(6,29, by=3) # taking every third segment column
df.mapping.table[, c(seg_col_range)] <- lapply(df.mapping.table[, c(seg_col_range)], function(x) round((x / as.numeric(df.mapping.table$`reads mapped`)*100), digits = 2))

df.mapping.table$`reads in` <- f.color_bar("lightgreen")(df.mapping.table$`reads in`)
df.mapping.table$`reads mapped` <- f.color_bar("lightgreen")(df.mapping.table$`reads mapped`)

write.csv(  x=df.mapping.table,
            row.names = FALSE,
            file = file.path(params$list_folder, "mapping_statistics.csv")
)

kbl(df.mapping.table,
    digits = 3,
    escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), fixed_thead = T) %>%
  scroll_box(height = "400px")
```


```{r bamstats_cleanup}
rm(df.bamstat.data)
rm(dt.bamstats.coverage)
rm(df.output)
rm(df.mapping.table)
```

### Coverage Distribution

The following plots show the read coverage of each sample after mapping. In 100 bp steps the mean read depth was calculated and plotted. Please be aware of the varying x and y axis scaling of the samples. The dotted read line shows the minimal required read depth.

```{r read_coverage}
# read file content
# dt.coverage <- as.data.table(ldply(l.infiles.coverage, fread, sep='\t'))
# colnames(dt.coverage) <- c("sample","chromosome", "position", "depth")

# reduce amount of data points to be plotted
dt.coverage[, bin:=rep(seq(1, ceiling(length(position) / 100)), each = 100, length.out = length(position)), by = "sample"]
dt.coverage[, mid.bin:=seq(1,length(position)) %% 100 ]
dt.coverage[, mean.cov:=mean(depth), by=c("sample", "bin")]

# adding placeholder row(s) for missing segment(s)
ls.segment <- list("*HA","*NA","*MP","*NP","*NS","*PA","*PB1","*PB2")

for (s in unique(dt.coverage$sample)){
  tmp <- dt.coverage[dt.coverage$sample == as.character(s),]
  for (l in ls.segment){
    if (Reduce("|",grepl(l, unique(tmp$chromosome)))) {}
    else {
      new_row <- list(s, substrRight(l,2), 0, NA, NA, 50, 1)
      dt.coverage <- rbind(dt.coverage, new_row)
    }
  }
  rm(tmp)
}

dt.coverage <- dt.coverage %>% group_by(chromosome)
dt.coverage <- dt.coverage %>% arrange(factor(chromosome, levels = c("HA","NA","MP","NP","NS","PA","PB1","PB2")))
dt.coverage <- dt.coverage %>% group_by(sample)
dt.coverage$chromosome = factor(dt.coverage$chromosome, levels=c("HA","NA","MP","NP","NS","PA","PB1","PB2"))
```


#### Sequence depth distribution on reference genome

Sequence depth was calculated at each position and plotted. The aim for positive samples is an evenly distributed high sequence depth.
The plot shows the sequence depth distribution on each segment. The positions were divided into bins with size of 100. In each bin the mean of the depths were then calculated and plotted. Please be aware of the varying x axis scaling.

<div class=superhighimage>

```{r plot_coverage, fig.width=10, fig.height=3}
# https://stackoverflow.com/questions/39119917/how-to-add-a-legend-to-hline
sample.names <- unique(dt.coverage$sample)
 for (s in c(sample.names)){
   plt <- ggplot(dt.coverage[(dt.coverage$sample == s) & (dt.coverage$mid.bin == 50),],
         aes(x=position, y=mean.cov, fill = chromosome, colour = chromosome)) +
         labs(title = paste("coverage distribution of", s, sep = " "), y = "mean coverage",
               fill = "chromosome") +
         geom_bar(stat = "identity", width = 1) +
         facet_grid(~chromosome, switch = 'y', scales = "free_x") +
         scale_x_continuous(limits = c(0,NA)) +
         theme(axis.text.x = element_text(angle = 45, size = 10, hjust = 1)) +
         geom_hline(aes(yintercept=min.coverage,
               linetype = paste("minimal coverage = ",as.character(min.coverage))),
               colour = "red", linewidth = 0.5) +
         scale_linetype_manual(name = "threshold", values = c(2),
               guide = guide_legend(override.aes = list(color = c("red")))) +
         theme(legend.key.size = unit(0.5, 'cm'), legend.title = element_text(size = 10),
               legend.text = element_text(size=8))
    print(plt)
 }
```

```{r coverage_cleanup}
rm(dt.coverage)
```

</div>
### Consensus Sequenz 
The consensus sequence was build using BCFtools. Only indels with a frequency above 90 %  were integrated into the consensus sequence. Positions with a coverage below the given threshold were masked with an N in the consensus sequence. Variants that passed the filtering steps were explicitly integrated into the consensus sequence if they occurred with a frequency of 90 % or higher. Variants with a frequency between 10 and 90 % were depicted as ambiguous bases and below 10 % the reference base was used. High quality Variants at the beginning or end of a genome segment, with an allele frequency above 90 % that failed the strand-bias filtering step were masked with N.

`r if(!reference){"Reads are compared against reference sequence databases for each segment during the automatic reference detection step . The following table lists the five best references per segment. On the first position is the reference that was used for the mapping."}`

``` {r read_ranking_top5_refs}

if (!reference) {
  # read file content
  dt.top.refs <- as.data.table(ldply(l.infiles.read.ranking.refs, fread, sep=' ', 
                                     select = c(1,16,17), na.strings = ""))
  colnames(dt.top.refs) <- c("sample","taxid", "magic number", "segment")
  
  dt.top.refs$taxid <- substr(dt.top.refs$taxid, 20, nchar(dt.top.refs$taxid)-3)
  dt.top.refs$sample <- substring(dt.top.refs$sample, 19)
  dt.top.refs <- dt.top.refs %>% group_by(sample, segment) %>% arrange(desc(`magic number`))
  dt.top.refs <- dt.top.refs %>% arrange(sample, segment)
  dt.top.refs <- dt.top.refs[,-3]
  
  dt.top.refs.na.filled <- data.table(sample = as.character(), taxid = as.character(), segment = as.character())
  
  # add placeholder for missing top ref 
  for (smpl in unique(dt.top.refs$sample)) {
    for (sgmnt in unique(dt.top.refs$segment)) {
      dt.group <- dt.top.refs[which( dt.top.refs$sample == smpl & dt.top.refs$segment == sgmnt) ,]
      while (nrow(dt.group)<5) {
        dt.group = rbind(dt.group, c(sample = smpl, taxid = "n/a", segment = sgmnt))
      }
      dt.top.refs.na.filled <- rbind(dt.top.refs.na.filled, dt.group)
      rm(dt.group)
    }
  }
  
  df.top.refs.reshaped <- tibble('sample' = character(),
                                 'reference hit' = numeric(),
                                 'taxid_HA' = character(),
                                 'taxid_NA' = character(),
                                 'taxid_MP' = character(),
                                 'taxid_NP' = character(),
                                 'taxid_NS' = character(),
                                 'taxid_PA' = character(),
                                 'taxid_PB1' = character(),
                                 'taxid_PB2' = character())
  
  for (s in c(unique(dt.top.refs.na.filled$sample))) {
    tmp <- dt.top.refs.na.filled[dt.top.refs.na.filled$sample == s,]
    tmp.df <- tibble('sample' = head(tmp$sample,5),
                     'reference hit' = 1:5,
                     'taxid_HA' = na.omit(tmp[tmp$segment == 'HA',]$taxid),
                     'taxid_NA' = tmp[tmp$segment == 'NA',]$taxid,
                     'taxid_MP' = na.omit(tmp[tmp$segment == 'MP',]$taxid),
                     'taxid_NP' = na.omit(tmp[tmp$segment == 'NP',]$taxid),
                     'taxid_NS' = na.omit(tmp[tmp$segment == 'NS',]$taxid),
                     'taxid_PA' = na.omit(tmp[tmp$segment == 'PA',]$taxid),
                     'taxid_PB1' = na.omit(tmp[tmp$segment == 'PB1',]$taxid),
                     'taxid_PB2' = na.omit(tmp[tmp$segment == 'PB2',]$taxid))
    df.top.refs.reshaped = rbind(df.top.refs.reshaped, tmp.df)
  }
  rm(tmp, tmp.df)
  write.csv(  x=df.top.refs.reshaped,
            row.names = FALSE,
            file = file.path(params$list_folder, "top5_consensus_sequence.csv")
  )
}
```


```{r table_ranking_top5_reference}
if (!reference) {
  # Format table
  df.top.refs.reshaped %>% 
    group_by(sample) %>% 
    slice(1:5) %>% 
    select(sample, everything()) %>% 
    collapse_rows_df(sample) %>%
    formattable() %>%
    kbl(digits = 3, col.names = NULL, align = c("r"), format = "html",
        escape = T, linesep = NA) %>%
    kable_styling(bootstrap_options = c("striped", "hover"), fixed_thead = T,
                  full_width = T) %>%
    add_header_above(c("Sample","Reference Hit","HA"=1,"NA"=1,"MP"=1,"NP"=1,
                       "NS"=1,"PA"=1,"PB1"=1,"PB2"=1)) %>%
    row_spec(seq(1,nrow(df.top.refs.reshaped),5), background="#EEEEEE") %>% 
    column_spec(1, color = "black" , background = "#EEEEEE")  %>%
    column_spec(3:10, width_min = '5in')  %>%
    scroll_box(height = "400px")
}
```

```{r top_refs_cleanup}
rm(dt.top.refs)
rm(dt.top.refs.na.filled)
rm(df.top.refs.reshaped)
```