Snakemake checkpoint报错：GATK染色体拆分步骤非零退出码

Let's break down why your scattering checkpoint is failing in Snakemake even though the command works manually, and fix it step by step:

1. Fix the Typo in Your Aggregation Function #

First, there's a typo in aggregate_scatter that will break downstream steps (even if it's not the root cause of your current error):

def aggregate_scatter(wildcards):
    checkpoint_output = checkpoints.scattering.get(**wildcards).output[0]
    # Changed checkoint_output → checkpoint_output below
    return expand(
        'scatter_{chr}/{i}/scattered.interval_list',
        chr=wildcards.chr,
        i=glob_wildcards(os.path.join(checkpoint_output, '{i}/scattered.interval_list')).i
    )

This typo would prevent Snakemake from finding the scattered files once the checkpoint runs successfully, so fix it first.

2. Diagnose the Exact Command Execution #

Run Snakemake with verbose logging to see the exact command being executed and any hidden error output:

snakemake --printshellcmds --verbose

This will show you:

• If the command Snakemake runs matches what you tested manually
• Any stderr/stdout output from GATK that Snakemake might be interpreting as an error (even if the exit code is 0)

3. Fix Shell Command Formatting & Output Handling #

Your shell command has minor formatting issues, and GATK might be writing non-error messages to stderr which Snakemake misinterprets. Rewrite the shell block to use triple quotes for readability and redirect output to a log file:

checkpoint scattering:
    input:
        interval = 'gcfiltered_{chr}.interval_list'
    output:
        directory('scatter_{chr}')
    shell:
        """
        mkdir -p {output} && \
        gatk --java-options "-Xmx8G" IntervalListTools \
            --INPUT {input.interval} \
            --SUBDIVISION_MODE INTERVAL_COUNT \
            --SCATTER_CONTENT 600 \
            --OUTPUT {output} > {output}/scatter.log 2>&1
        """

The 2>&1 redirects stderr to stdout, which ensures Snakemake doesn't flag normal GATK logging as an error. The log file will also help you debug if something goes wrong.

4. Address Snakemake Version Limitations #

You're using Snakemake 5.5.4, a very old version (released in 2019). The checkpoint and directory output handling have been significantly improved in newer versions—older Snakemake versions sometimes had bugs detecting successful completion of commands that output directories.

Upgrade to a recent stable version:

pip install --upgrade snakemake>=7.0.0

If you can't upgrade for environment reasons, add a "done" flag file to the checkpoint output to explicitly signal completion:

checkpoint scattering:
    input:
        interval = 'gcfiltered_{chr}.interval_list'
    output:
        directory('scatter_{chr}'),
        'scatter_{chr}/.done'
    shell:
        """
        mkdir -p {output[0]} && \
        gatk --java-options "-Xmx8G" IntervalListTools \
            --INPUT {input.interval} \
            --SUBDIVISION_MODE INTERVAL_COUNT \
            --SCATTER_CONTENT 600 \
            --OUTPUT {output[0]} > {output[0]}/scatter.log 2>&1 && \
        touch {output[1]}
        """

The .done file tells Snakemake to wait until the GATK command fully completes before considering the checkpoint done.

5. Verify Input File Compatibility #

Double-check that your gcfiltered_{chr}.interval_list file is properly formatted for GATK 4.1.2.0. Your sample input looks correct, but ensure there are no trailing spaces or invalid lines that might cause GATK to silently fail (even if the exit code is 0).

Final Testing #

After applying these fixes, run Snakemake again. If the checkpoint still fails, check the scatter_{chr}/scatter.log file for any GATK-specific warnings or errors that might be causing the issue.

内容的提问来源于stack exchange，提问作者dmkr