view data2.R @ 2:e089fb4ee28b draft

planemo upload for repository https://github.com/bernt-matthias/mb-galaxy-tools/tree/topic/dada2/tools/dada2 commit 5b1603bbcd3f139cad5c876be83fcb39697b5613-dirty

library(dada2)

# library(DBI)
# library(ggplot2)
library(optparse)
# library(RSQLite)
# library(stringr)

## source required R functions
source('user_input_functions.R')

# print dada2 version
print(paste("dada2 version: ", packageVersion("dada2")))

# # R function to create fasta file from dada2 output data
# outdir is directory to output fasta file
# taxa is taxonomy file generated by dada2
# prefix is string for desired prefix attached to output file names

dada2fasta <- function(outdir, seqtab.nochim, prefix){
  seq <- colnames(seqtab.nochim)
  n <- 0
  ASVs <- c()
  for(i in seq){
    n <- n + 1
    ASV <- paste('ASV', as.character(n), sep = '_')
    ASVs <- c(ASVs, ASV)
    line1 <- paste('>',ASV,sep='')
    write(line1, file.path(outdir,sprintf('%s.fasta',prefix)), append=T)
    write(i, file.path(outdir,sprintf('%s.fasta',prefix)), append=T)
  }
  return(ASVs)
}


# # R DADA2 workflow
# wd is path to fastq files
# r_path is path to user_input_functions.R
# outdir is path to output directory 
# prefix is string for desired prefix attached to output file names

dada2run <- function(wd, r_path, outdir, prefix){
  
  # read-in files-------------------------------------------------------
  ## obtain vectors of forward and reverse reads based on 'R1' and 'R2' in file names
  ## additionally obtain the coressponding sample names for these files
  p1 <- c()
  p2 <- c()
  sample.names <- c()
  for(f in list.files(wd, full.names=T)){
    if(grepl('_1.fq', f)){
      sample <- gsub('^.*[/](.*?)_1\\.fq\\.gz', '\\1', f)
      sample.names <- c(sample.names, sample)
      p1 <- c(p1, f)
    }
    if(grepl('_2.fq', f)){
      p2 <- c(p2, f)
    }
  }
  fnFs <- sort(p1)
  fnRs <- sort(p2)
  sample.names <- sort(sample.names)

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_test.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_test.Rdata')))
  
  ## print for review
  to_view <- data.frame(sample.names, fnFs, fnRs)
  cat("The following fastq files were found:\n")
  print(to_view)
  
  # Perform quality filtering and trimming---------------------------------
  ## assign new names to samples
  filtFs <- file.path(outdir, paste0(sample.names, 'F_filt.fastq.gz'))
  filtRs <- file.path(outdir, paste0(sample.names, 'R_filt.fastq.gz'))
  
  ## plot forward and reverse quality so that user can decide on filtering parameters
  cat('Plotting quality profile of forward reads...\n')
  Fqp1 <- plotQualityProfile(fnFs[1])
  #print(Fqp1)
  ggsave(sprintf('%s_forward_1_qualityprofile.pdf',prefix), Fqp1, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_forward_1_qualityprofile.emf',prefix), Fqp1, path = outdir, width = 20,height = 15,units = c("cm"))
  Fqp2 <- plotQualityProfile(fnFs[2])
  #print(Fqp2)
  ggsave(sprintf('%s_forward_2_qualityprofile.pdf',prefix),Fqp2, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_forward_2_qualityprofile.emf',prefix), Fqp2, path = outdir, width = 20,height = 15,units = c("cm"))
  #cat('Which position would you like to truncate the forward reads at?\nPlease use the red-dashed lines as a guide, where they stop appearing indicates good quality.\nNOTE: Do NOT over-trim! You still require overlap between your forward and reverse reads in order to merge them later!\n')
  len1 <- 240
  cat('Plotting quality profile of reverse reads...\n')
  Rqp1 <- plotQualityProfile(fnRs[1])
  #print(Rqp1)
  ggsave(sprintf('%s_reverse_1_qualityprofile.pdf',prefix),Rqp1, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_reverse_1_qualityprofile.emf',prefix), Rqp1, path = outdir, width = 20,height = 15,units = c("cm"))
  Rqp2 <- plotQualityProfile(fnRs[2])
  #print(Rqp2)
  ggsave(sprintf('%s_reverse_2_qualityprofile.pdf',prefix), Rqp2, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_reverse_2_qualityprofile.emf',prefix), Rqp2, path = outdir, width = 20,height = 15,units = c("cm"))
  #cat('Which position would you like to truncate the forward reads at?\nPlease use the red-dashed lines as a guide, where they stop appearing indicates good quality.\nNOTE: Do NOT over-trim! You still require overlap between your forward and reverse reads in order to merge them later!\n')
  len2 <- 240

  ## filter and trim
  ## remaining parameters set to recommended defaults
  ## maxN must be set to 0 (DADA2 requries no Ns)
  ## The maxEE parameter sets the maximum number of "expected errors" allowed in a read, which is a better filter than simply averaging quality scores.
  ## If not using Windows, you may set multithread to TRUE
  ## NOTE: Do not use the truncLen parameter for ITS sequencing
  ## trimLeft needs to be varied based on the size of your primers (i.e. it is used to trim your primer sequences)!! 
  cat('Filtering and trimming sequences...\n')
  out <- filterAndTrim(fnFs, filtFs, fnRs, filtRs, truncLen=c(len1,len2), maxN=0, maxEE=c(2,2), truncQ=2, rm.phix=T, compress=T, multithread=threads, trimLeft=15)
  
  ## have user review read count changes, and relax error rate if too many reads are lost
  ## for example, you may especially want to relax the number of expected errors on the reverse reads (i.e. maxEE=c(2,5)), as the reverse is prone to error on the Illumina sequencing platform
  print(head(out))
  check2 <- T
  while(check2 == F){
    maxF <- numeric_input("What would you like the maximum number of expected errors in the forward reads to be?\nDefault 2:", 2)
    maxR <- numeric_input("What would you like the maximum number of expected errors in the reverse reads to be?\nDefault 5:", 5)
    out <- filterAndTrim(fnFs, filtFs, fnRs, filtRs, truncLen=c(len1,len2), maxN=0, maxEE=c(maxF,maxR), truncQ=2, rm.phix=T, compress=T, multithread=threads)
    print(head(out))
    check2 <- yn_input('Proceed? If you lost too many reads, you can choose to not proceed and you will have the option to relax the error rate. Default yes:',T)
  }
  
  # Have DADA2 learn the error rates-----------------------------------------------
  ## If not using Windows, you may set multithread to TRUE
  read.subset <- 1e6
  cat('Learning error rate of forward reads...\n')
  errF <- learnErrors(filtFs, nreads=read.subset, multithread=threads)
  
  ## have user check estimated error rates
  ## note the calculations are done with a subset of the total reads, as it is computationally intensive
  ## if the fit is poor, the user has the option to try again with an increased subset number of reads
  Error_f <- plotErrors(errF, nominalQ = T)
  #print(Error_f)
  ggsave(sprintf('%s_forward_Error_plot.pdf',prefix), Error_f, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_forward_Error_plot.emf',prefix), Error_f, path = outdir, width = 20,height = 15,units = c("cm"))
  check3a <- T
  while(check3a == F){
    read.subset <- numeric_input('Please specify the number of reads you would like dada2 to utilize to calculate the error rate.\nThe default previously used was 1e6.\nThe newly recommended default is 10-fold greater,\n1e7:',1e7)
    errF <- learnErrors(filtFs, nreads=read.subset, multithread=threads)
    print(Error_f)
    ggsave(sprintf('%s_forward_Error_plot.pdf',prefix), path = outdir, width = 20,height = 15,units = c("cm"))
    ggsave(sprintf('%s_forward_Error_plot.emf',prefix), path = outdir, width = 20,height = 15,units = c("cm"))
    check3a <- yn_input('Proceed?\nThe estimated error rate (black line) should fit to the observed error rates for each consensus quality score (black points).\nAdditionally, the error rates expected under the nominal definition of the Q-value (quality score) should decrease as the quality score increases (or flat-line).\nIf you do not have a good fit, you may want dada2 to re-learn the error rates with a higher number of reads in the utilized subset.\nA subset of reads is always used as the algorithm is computationally intensive.\nDefault yes:',T)
  }
  

  ## also do for reverseL
  cat('Learning error rate of reverse reads...\n')
  errR <- learnErrors(filtRs, nreads=read.subset, multithread=threads)
  Error_r <- plotErrors(errR, nominalQ=T)
  #print(Error_r)
  ggsave(sprintf('%s_reverse_Error_plot.pdf',prefix), Error_r, path = outdir, width = 20,height = 15,units = c("cm"))
  #ggsave(sprintf('%s_reverse_Error_plot.emf',prefix), Error_r, path = outdir, width = 20,height = 15,units = c("cm"))
  check3b <- T
  while(check3b == F){
    read.subset <- numeric_input('Please specify the number of reads you would like dada2 to utilize to calculate the error rate.\nThe default previously used was 1e6.\nThe newly recommended default is 10-fold greater,\n1e7:',1e7)
    errR <- learnErrors(filtRs, nreads=read.subset, multithread=threads)
    print(Error_r)
    ggsave(sprintf('%s_reverse_Error_plot.pdf',prefix), path = outdir, width = 20,height = 15,units = c("cm"))
    #ggsave(sprintf('%s_reverse_Error_plot.emf',prefix), path = outdir, width = 20,height = 15,units = c("cm"))
    check3b <- yn_input('Proceed?\nThe estimated error rate (black line) should fit to the observed error rates for each consensus quality score (black points).\nAdditionally, the error rates expected under the nominal definition of the Q-value (quality score) should decrease as the quality score increases (or flat-line).\nIf you do not have a good fit, you may want dada2 to re-learn the error rates with a higher number of reads in the utilized subset.\nA subset of reads is always used as the algorithm is computationally intensive.\nDefault yes:',T)
  }

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_learning.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_post_learning.Rdata')))
  
  # Dereplicate sequences to generate unique sequence fastq files with corresponding count tables-------------------------
  ## NOTE: if your dataset is huge, you may run out of RAM. Please see https://benjjneb.github.io/dada2/bigdata.html for details.
  cat('Dereplicating forward reads...\n')
  derepFs <- derepFastq(filtFs, verbose=T)
  cat('Dereplicating reverse reads...\n')
  derepRs <- derepFastq(filtRs, verbose=T)
  
  ## name the derep-class objects by sample names
  names(derepFs) <- sample.names
  names(derepRs) <- sample.names

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_derep.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_post_derep.Rdata')))
  
  # Infer sequence variants using learned error rates---------------------------------
  ## If not using Windows, you may set multithread to TRUE
  ## NOTE: if your dataset is huge, you may run out of RAM. Please see https://benjjneb.github.io/dada2/bigdata.html for details.
  ## NOTE2: you can use DADA2 for 454 or IonTorrent data as well. Please see https://benjjneb.github.io/dada2/tutorial.html.
  s.pool = F
  cat('Inferring sequence variants of forward reads...\n')
  dadaFs <- dada(derepFs, err=errF, pool=s.pool, multithread=threads)
  
  ## have user inspect detected number of sequence variants, to ensure it is logical based on the biological context of their samples
  ## if you have low sampling depths, you may not want to process each sample independently as per default, but set pool=T. It gives better results at increased computation time. The user will have the option to do this if the number of sequence variants doesn't make sense.
  print(dadaFs[[1]])
  check4 <- T
  if(check4 == F){
    s.pool = T
    dadaFs <- dada(derepFs, err=errF, pool=s.pool, multithread=threads)
    print(dadaFs[[1]])
    cat('Hopefully, these results make more sense.\nOtherwise, there is not much more you can do except start over!\n')
    check <- yn_input('Proceed? Default yes, no to quit:',T)
    if(check == F){
      stop()
    }
  }
  
  ## also do for reverse, but don't need to re-check as you need to keep the pool consistent between the forward and reverse!
  cat('Inferring sequence variants of reversed reads...\n')
  dadaRs <- dada(derepRs, err=errR, pool=s.pool, multithread=threads)

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_dada.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_post_dada.Rdata')))

  
  # Merge forward and reverse reads-------------------------------------------------
  cat('Merging paired-end reads...\n')
  mergers <- mergePairs(dadaFs, derepFs, dadaRs, derepRs, verbose=T)
  #cat('Most of your reads should have been retained (i.e. were able to merge, see above).\nOtherwise, there is not much more you can do except start over (i.e. did you over-trim your sequences??)!\n')
  check <- T
  if(check == F){
    stop()
  }
  
  # Construct sequences table-------------------------------------------------------
  cat('Constructing sequence table...\n')
  seqtab <- makeSequenceTable(mergers)

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_merge.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_post_merge.Rdata')))


  ## inspect distribution of sequence lengths
  ## give user the option to filter out overly long or short sequneces
  cat('Sequence length distribution listed below:\n')
  print(table(nchar(getSequences(seqtab))))
  check5 <- T
  if(check5 == F){
    min.cutoff <- numeric_input('Please input desired minimum length of sequences:',NULL)
    max.cutoff <- numeric_input('Please input desired maximum length of sequences:',NULL)
    seqtab <- seqtab[,nchar(colnames(seqtab)) %in% seq(min.cutoff,max.cutoff)]
  }
  
  # Remove chimeras------------------------------------------------------------------
  ## If not using Windows, you may set multithread to TRUE
  cat('Removing chimeras...\n')
  seqtab.nochim <- removeBimeraDenovo(seqtab, method='consensus', multithread=threads, verbose=T)
  
  ## display percentage of chimeras removed
  ## this number should be small (<5%), otherwise some processing parameters need to be revisited
  percent.nochim <- (sum(seqtab.nochim)/sum(seqtab))*100
  percent.nochim <- paste(as.character(percent.nochim),'of reads retained after chimera removal.\n',sep=' ')
  cat(percent.nochim)
  #cat('Most of your reads should have been retained.\nOtherwise, there is not much more you can do except start over!\n')
  check <- T
  if(check == F){
    stop()
  }
  
  # Final sanity check--------------------------------------------------------------
  ## track reads removed throughout the pipeline
  ## If processing a single sample, remove the sapply calls: e.g. replace sapply(dadaFs, getN) with getN(dadaFs)
  getN <- function(x) sum(getUniques(x))
  track <- cbind(out, sapply(dadaFs, getN), sapply(mergers, getN), rowSums(seqtab), rowSums(seqtab.nochim))
  colnames(track) <- c("input", "filtered", "denoised", "merged", "tabled", "nonchim")
  rownames(track) <- sample.names
  print(head(track))
  #cat('Most of your reads should have been retained.\nOtherwise, there is not much more you can do except start over!\n')
  check <- T
  if(check == F){
    stop()
  }
  
  write.csv(track,file=file.path(outdir, sprintf('%s_read_count-quality_control.csv',prefix)))

  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_chimera.Rdata')))
#   load(file = file.path(outdir, paste0(prefix, 'state_post_chimera.Rdata')))
  
  # Assign taxonomy-----------------------------------------------------------------
  ## require silva database files
  ## If not using Windows, you may set multithread to TRUE
  ## Minimum boot strap should be 80, but for sequnce length =< 250 Minimum bootstrap set to 50 (which is also the default)
  
  ## SILVA
  cat('Assigning taxonomy to genus level using SILVA...\n')
  taxa_silva <- assignTaxonomy(seqtab.nochim, file.path(wd,"silva_nr_v132_train_set.fa.gz"), multithread=threads, minBoot=80, tryRC=T)
  cat('Assigning taxonomy at species level using SILVA...\n')
  taxa_silva <- addSpecies(taxa_silva, file.path(wd,"silva_species_assignment_v132.fa.gz"), allowMultiple=T, tryRC=T)
  write.csv(taxa_silva,file=file.path(outdir, sprintf('%s_taxonomy_silva.csv',prefix)))
  
  ## RDP - used for copy number correction
  cat('Assigning taxonomy to genus level using RDP...\n')
  taxa_rdp <- assignTaxonomy(seqtab.nochim, file.path(wd,"rdp_train_set_16.fa.gz"), multithread=threads, minBoot=80, tryRC=T)
  cat('Assigning taxonomy at species level using RDP...\n')
  taxa_rdp <- addSpecies(taxa_rdp, file.path(wd,"rdp_species_assignment_16.fa.gz"), allowMultiple=T, tryRC=T)
  write.csv(taxa_rdp,file=file.path(outdir, sprintf('%s_taxonomy_rdp.csv',prefix)))
  save(list = ls(all.names = TRUE), file = file.path(outdir, paste0(prefix, 'state_post_tax.Rdata')))
  #load(file = file.path(outdir, paste0(prefix, 'state_post_tax.Rdata')))


  # Return data----------------------------------------------------------------------
  cat('Returning data...\n')
  ## create fasta file
  ASVs <- dada2fasta(outdir, seqtab.nochim, prefix)
  ## create master dataframe for each classification
  
  ## Assigning ASVs to count table
  sequences <- colnames(seqtab.nochim)
  colnames(seqtab.nochim) <- ASVs
  seqtab.nochim <- t(seqtab.nochim)
  
  ## silva
  taxa_silva <- taxa_silva[match(sequences, rownames(taxa_silva)),]
  rownames(taxa_silva) <- ASVs
  d <- merge(seqtab.nochim, taxa_silva, by='row.names')
  colnames(d)[1] <- 'ASV'
  ## create database of all information
  db <- dbConnect(RSQLite::SQLite(), file.path(outdir, sprintf('%s.sqlite',prefix)))
  dbWriteTable(db, 'dada2_results_silva', d)
  ## write master dataframe for user, and return it
  write.table(d, file.path(outdir, sprintf('%s_dada2_results_silva.txt',prefix)), sep='\t', quote=F, row.names=F)
  
  ## rdp
  taxa_rdp <- taxa_rdp[match(sequences, rownames(taxa_rdp)),]
  rownames(taxa_rdp) <- ASVs
  d <- merge(seqtab.nochim, taxa_rdp, by='row.names')
  colnames(d)[1] <- 'ASV'
  ## create database of all information
  dbWriteTable(db, 'dada2_results_rdp', d)
  dbDisconnect(db)
  ## write master dataframe for user, and return it
  write.table(d, file.path(outdir, sprintf('%s_dada2_results_rdp.txt',prefix)), sep='\t', quote=F, row.names=F)
  return(d)
  
  cat('DADA2 processing completed!\n')
}

 
option_list = list(
  make_option(c("-t", "--threads"), type = "integer", default = 1,
              help = "number of threads to use", metavar = "THREADS")
); 

opt_parser = OptionParser(option_list=option_list);
opt = parse_args(opt_parser);

print(opt)

threads = as.integer(Sys.getenv("NSLOTS", "1"))

exit(1)



dada2run(wd='/work/haange/Leaky_gut/', r_path='/work/haange/Leaky_gut', outdir='/work/haange/Leaky_gut/results', prefix='Leaky_gut')