GATK MARKDUPLICATESSPARK

Spark implementation of Picard MarkDuplicates that allows the tool to be run in parallel on multiple cores on a local machine or multiple machines on a Spark cluster while still matching the output of the non-Spark Picard version of the tool. Since the tool requires holding all of the readnames in memory while it groups read information, machine configuration and starting sort-order impact tool performance.

URL: https://gatk.broadinstitute.org/hc/en-us/articles/9570319741083-MarkDuplicatesSpark

Example

This wrapper can be used in the following way:

rule mark_duplicates_spark:
    input:
        "mapped/{sample}.bam",
    output:
        bam="dedup/{sample}.bam",
        metrics="dedup/{sample}.metrics.txt",
    log:
        "logs/dedup/{sample}.log",
    params:
        extra="--remove-sequencing-duplicates",  # optional
        java_opts="",  # optional
        #spark_runner="",  # optional, local by default
        #spark_v4.6.0-24-g250dd3e="",  # optional
        #spark_extra="", # optional
    resources:
        # Memory needs to be at least 471859200 for Spark, so 589824000 when
        # accounting for default JVM overhead of 20%. We round round to 650M.
        mem_mb=lambda wildcards, input: max([input.size_mb * 0.25, 650]),
    threads: 8
    wrapper:
        "v4.6.0-24-g250dd3e/bio/gatk/markduplicatesspark"

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.

Notes

• The java_opts param allows for additional arguments to be passed to the java compiler, e.g. “-XX:ParallelGCThreads=10” (not for -XmX or -Djava.io.tmpdir, since they are handled automatically).

• The extra param allows for additional program arguments.

• The spark_runner param = “LOCAL”|”SPARK”|”GCS” allows to set the spark_runner. Set the parameter to “LOCAL” or don’t set it at all to run on local machine.

• The spark_master param allows to set the URL of the Spark Master to submit the job. Set to “local[number_of_cores]” for local execution. Don’t set it at all for local execution with number of cores determined by snakemake.

• The spark_extra param allows for additional spark arguments.

Software dependencies

• gatk4=4.5.0.0

• snakemake-wrapper-utils=0.6.2

Input/Output

Input:

• bam file

• reference file

Output:

• bam file with marked or removed duplicates

Authors

• Filipe G. Vieira

Code

__author__ = "Fillipe G. Vieira"
__copyright__ = "Copyright 2021, Filipe G. Vieira"
__license__ = "MIT"

import tempfile
from snakemake.shell import shell
from snakemake_wrapper_utils.java import get_java_opts

extra = snakemake.params.get("extra", "")
spark_runner = snakemake.params.get("spark_runner", "LOCAL")
spark_master = snakemake.params.get(
    "spark_master", "local[{}]".format(snakemake.threads)
)
spark_extra = snakemake.params.get("spark_extra", "")
java_opts = get_java_opts(snakemake)

metrics = snakemake.output.get("metrics", "")
if metrics:
    metrics = f"--metrics-file {metrics}"

log = snakemake.log_fmt_shell(stdout=True, stderr=True)

with tempfile.TemporaryDirectory() as tmpdir:
    shell(
        "gatk --java-options '{java_opts}' MarkDuplicatesSpark"
        " --input {snakemake.input}"
        " {extra}"
        " --tmp-dir {tmpdir}"
        " --output {snakemake.output.bam}"
        " {metrics}"
        " -- --spark-runner {spark_runner} --spark-master {spark_master} {spark_extra}"
        " {log}"
    )