ont-open-data-dataset/s3://ont-open-data/gm24385_2020.09/config/setup_katuali.sh

#!/bin/bash

# Set up katuali workspace for multiple runfolders:
#  1. create top-level folder
#  2. copy sequencing_summary.txt
#  3. concatenate all fastqs to guppy_<suffix>/basecalls.fastq
#
# Finally
#  4. symlink the ref directory alongside data directories

# Usage:
# run from location where data should be placed
# setup_katuali.sh <guppy_suffix> ../../runfolder1 ../runfolder2 ...

guppy_suffix=$1
shift

# directory of this script, used to locate other resources
DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"

for folder in $@; do
    echo "Creating workspace for: $folder"
    name=$(basename $folder)
    abspath=$(realpath $folder)
    echo "name: $name"
    echo "source: $abspath"
    echo $name
    mkdir -p $name

    # symlink fast5
    fast5_src=$abspath/fast5_pass
    if [[ -d "$fast5_src" ]]; then
        echo "Symlinking fast5_pass to $name/reads"
        ln -s ${fast5_src} $name/reads
    else
        echo "Expected fast5 location does not exist: ${fast5_src}"
        exit 1
    fi

    # concatenate fastq
    fastq_src=$abspath/fastq_pass
    guppy_out=${name}/guppy_${guppy_suffix}
    if [[ -d "$fastq_src" ]]; then
        echo "Collating fastq_pass to ${guppy_out}" 
        mkdir -p "${guppy_out}"
        ${DIR}/rationalize_fastq -i ${fastq_src} -o ${guppy_out}/basecalls.fastq.gz
    else
        echo "Expected fast5 location does not exist: ${fastq_src}"
        exit 1
    fi

    # copy summary
    summary=$(ls ${abspath}/sequencing_summary_????????_????????.txt 2>/dev/null)
    summary_out=${guppy_out}/sequencing_summary.txt
    echo $summary
    if [[ -n "${summary}" ]]; then
        echo "Copying ${summary} to ${summary_out}"
        cp ${summary} ${summary_out}
    else
        echo "Could not find sequencing summary in ${abspath}"
        exit 1
    fi
done

# symlink ref directory
ref_dir=$(realpath ${DIR}/ref)
ln -s ${ref_dir} ref