DADA2_LEARN_ERRORS

DADA2 Learning error rates separately on paired-end data using dada2 learnErrors function. Optional parameters are documented in the manual and the function is introduced in the dedicated tutorial section.

Example

This wrapper can be used in the following way:

rule learn_pe:
    # Run twice dada2_learn_errors: on forward and on reverse reads
    input: expand("results/dada2/model_{orientation}.RDS", orientation=[1,2])

rule dada2_learn_errors:
    input:
    # Quality filtered and trimmed forward FASTQ files (potentially compressed)
        expand("filtered/{sample}.{{orientation}}.fastq.gz", sample=["a","b"])
    output:
        err="results/dada2/model_{orientation}.RDS",# save the error model
        plot="reports/dada2/errors_{orientation}.png",# plot observed and estimated rates
    # Even though this is an R wrapper, use named arguments in Python syntax
    # here, to specify extra parameters. Python booleans (`arg1=True`, `arg2=False`)
    # and lists (`list_arg=[]`) are automatically converted to R.
    # For a named list as an extra named argument, use a python dict
    #   (`named_list={name1=arg1}`).
    #params:
    #    randomize=True
    log:
        "logs/dada2/learn-errors/learn-errors_{orientation}.log"
    threads: 1 # set desired number of threads here
    wrapper:
        "v3.9.0/bio/dada2/learn-errors"

Note that input, output and log file paths can be chosen freely.

When running with

snakemake --use-conda

the software dependencies will be automatically deployed into an isolated environment before execution.

Software dependencies

• bioconductor-dada2=1.30.0

Input/Output

Input:

• A list of quality filtered and trimmed forward FASTQ files (potentially compressed)

Output:

• err: RDS file with the stored error model

• plot: plot observed vs estimated errors rates

Params

• optional arguments for ``learnErrors(), please provide them as python key=value pairs``:

Authors

• Charlie Pauvert

Code

# __author__ = "Charlie Pauvert"
# __copyright__ = "Copyright 2020, Charlie Pauvert"
# __email__ = "cpauvert@protonmail.com"
# __license__ = "MIT"

# Snakemake wrapper for learning error rates on sequence data using dada2 learnErrors function.

# Sink the stderr and stdout to the snakemake log file
# https://stackoverflow.com/a/48173272
log.file<-file(snakemake@log[[1]],open="wt")
sink(log.file)
sink(log.file,type="message")

library(dada2)

# Prepare arguments (no matter the order)
args<-list(
           fls = snakemake@input,
           multithread=snakemake@threads
           )
# Check if extra params are passed
if(length(snakemake@params) > 0 ){
       # Keeping only the named elements of the list for do.call()
       extra<-snakemake@params[ names(snakemake@params) != "" ]
       # Add them to the list of arguments
       args<-c(args, extra)
} else{
    message("No optional parameters. Using defaults parameters from dada2::learnErrors()")
}

# Learn errors rates for both read types
err<-do.call(learnErrors, args)

# Plot estimated versus observed error rates to validate models
perr<-plotErrors(err, nominalQ = TRUE)

# Save the plots
library(ggplot2)
ggsave(snakemake@output[["plot"]], perr, width = 8, height = 8, dpi = 300)

# Store the estimated errors as RDS files
saveRDS(err, snakemake@output[["err"]],compress = T)

# Proper syntax to close the connection for the log file
# but could be optional for Snakemake wrapper
sink(type="message")
sink()