Examples with InputFileDescriptor com.hartwig.batch.input.InputFileDescriptor used on opensource projects

Example 26 with InputFileDescriptor

use of com.hartwig.batch.input.InputFileDescriptor in project pipeline5 by hartwigmedical.

the class RnaIsofoxSpliceJunctions method execute.

@Override
public VirtualMachineJobDefinition execute(InputBundle inputs, RuntimeBucket bucket, BashStartupScript startupScript, RuntimeFiles executionFlags) {
    InputFileDescriptor descriptor = inputs.get();
    final String batchInputs = descriptor.inputValue();
    final String[] batchItems = batchInputs.split(",");
    if (batchItems.length < 2) {
        System.out.print(String.format("invalid input arguments(%s) - expected SampleId,ReadLength", batchInputs));
        return null;
    }
    final String sampleId = batchItems[COL_SAMPLE_ID];
    final String geneIds = batchItems[COL_GENE_IDS];
    final RefGenomeVersion refGenomeVersion = V37;
    final ResourceFiles resourceFiles = buildResourceFiles(refGenomeVersion);
    final String samplesDir = String.format("%s/%s", getRnaCohortDirectory(refGenomeVersion), "samples");
    // copy down BAM and index file for this sample
    final String bamFile = String.format("%s%s", sampleId, RNA_BAM_FILE_ID);
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", samplesDir, sampleId, bamFile, VmDirectories.INPUT));
    final String bamIndexFile = String.format("%s%s", sampleId, RNA_BAM_INDEX_FILE_ID);
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", samplesDir, sampleId, bamIndexFile, VmDirectories.INPUT));
    // copy down the executable
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s %s", ISOFOX_LOCATION, ISOFOX_JAR, VmDirectories.TOOLS));
    startupScript.addCommand(() -> format("cd %s", VmDirectories.OUTPUT));
    // run Isofox
    StringJoiner isofoxArgs = new StringJoiner(" ");
    isofoxArgs.add(String.format("-sample %s", sampleId));
    isofoxArgs.add(String.format("-functions %s", FUNC_TRANSCRIPT_COUNTS));
    isofoxArgs.add(String.format("-output_dir %s/", VmDirectories.OUTPUT));
    isofoxArgs.add(String.format("-bam_file %s/%s", VmDirectories.INPUT, bamFile));
    isofoxArgs.add(String.format("-ref_genome %s", resourceFiles.refGenomeFile()));
    isofoxArgs.add(String.format("-ensembl_data_dir %s", resourceFiles.ensemblDataCache()));
    isofoxArgs.add(String.format("-write_exon_data"));
    isofoxArgs.add(String.format("-restricted_gene_ids %s", geneIds));
    isofoxArgs.add(" -output_id gene_sj");
    startupScript.addCommand(() -> format("java -jar %s/%s %s", VmDirectories.TOOLS, ISOFOX_JAR, isofoxArgs.toString()));
    // upload the results
    startupScript.addCommand(new OutputUpload(GoogleStorageLocation.of(bucket.name(), "isofox"), executionFlags));
    return ImmutableVirtualMachineJobDefinition.builder().name("rna-isofox").startupCommand(startupScript).namespacedResults(ResultsDirectory.defaultDirectory()).workingDiskSpaceGb(MAX_EXPECTED_BAM_SIZE_GB).build();
}

Example 27 with InputFileDescriptor

use of com.hartwig.batch.input.InputFileDescriptor in project pipeline5 by hartwigmedical.

the class RnaRsem method execute.

@Override
public VirtualMachineJobDefinition execute(InputBundle inputs, RuntimeBucket bucket, BashStartupScript startupScript, RuntimeFiles executionFlags) {
    InputFileDescriptor descriptor = inputs.get();
    final String batchInputs = descriptor.inputValue();
    final String[] batchItems = batchInputs.split(",");
    if (batchItems.length != 2) {
        System.out.print(String.format("invalid input arguments(%s) - expected SampleId,PathToFastqFiles", batchInputs));
        return null;
    }
    final String sampleId = batchItems[0];
    final String fastqFilelist = batchItems[1];
    final List<String> sampleFastqFiles = getSampleFastqFileList(sampleId, fastqFilelist);
    if (sampleFastqFiles.isEmpty()) {
        System.out.print(String.format("sampleId(%s) fastq files not found", sampleId));
        return null;
    }
    // copy down FASTQ files for this sample
    for (final String fastqFile : sampleFastqFiles) {
        startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s %s", fastqFile, VmDirectories.INPUT));
    }
    // download the executables
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp -r %s/%s %s", RSEM_RESOURCES, RSEM_TOOL, VmDirectories.TOOLS));
    startupScript.addCommand(() -> format("chmod a+x %s/%s/*", VmDirectories.TOOLS, RSEM_TOOL));
    // locate the FASTQ files for reads 1 and 2
    final String r1Files = format("$(ls %s/*_R1* | tr '\\n' ',')", VmDirectories.INPUT);
    final String r2Files = format("$(ls %s/*_R2* | tr '\\n' ',')", VmDirectories.INPUT);
    // download reference files
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp -r %s/%s %s", RNA_RESOURCES, REF_GENCODE_37_DIR, VmDirectories.INPUT));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp -r %s/%s %s", RSEM_RESOURCES, RSEM_GENE_INDEX_DIR, VmDirectories.INPUT));
    startupScript.addCommand(() -> format("cd %s", VmDirectories.OUTPUT));
    // run STAR with transcriptome mapping
    final String refGenomeDir = String.format("%s/%s", VmDirectories.INPUT, REF_GENCODE_37_DIR);
    final String threadCount = Bash.allCpus();
    final String[] starArgs = { "--runThreadN", threadCount, "--genomeDir", refGenomeDir, "--genomeLoad", "NoSharedMemory", "--readFilesIn", r1Files, r2Files, "--readFilesCommand", "zcat", "--outSAMtype", "BAM", "Unsorted", "--outSAMunmapped", "Within", "--outBAMcompression", "0", "--outSAMattributes", "All", "--outFilterMultimapNmax", "10", "--outFilterMismatchNmax", "3", "limitOutSJcollapsed", "3000000", "--chimSegmentMin", "10", "--chimOutType", "WithinBAM", "SoftClip", "--chimJunctionOverhangMin", "10", "--chimSegmentReadGapMax", "3", "--chimScoreMin", "1", "--chimScoreDropMax", "30", "--chimScoreJunctionNonGTAG", "0", "--chimScoreSeparation", "1", "--outFilterScoreMinOverLread", "0.33", "--outFilterMatchNminOverLread", "0.33", "--outFilterMatchNmin", "35", "--alignSplicedMateMapLminOverLmate", "0.33", "--alignSplicedMateMapLmin", "35", "--alignSJstitchMismatchNmax", "5", "-1", "5", "5", "--quantMode", // key line for RSEM;
    "TranscriptomeSAM" };
    startupScript.addCommand(new VersionedToolCommand("star", "STAR", "2.7.3a", starArgs));
    // key output file is 'Aligned.toTranscriptome.out.bam'
    // ./tools/RSEM-1.3.3/rsem-calculate-expression --alignments --paired-end
    // ./runs/CPCT02020378T/Aligned.toTranscriptome.out.bam
    // ./ref/rsem_gene_index/human_gencode
    // CPCT02020378T.rsem -p 6 &
    final String transcriptomeBam = "Aligned.toTranscriptome.out.bam";
    // TMP: copy transcriptome BAM to the bucket
    // startupScript.addCommand(() -> format("gsutil -m cp %s/%s gs://rna-cohort/%s/rsem/", VmDirectories.OUTPUT, transcriptomeBam, sampleId));
    final String rsemGeneIndex = String.format("%s/%s/%s", VmDirectories.INPUT, RSEM_GENE_INDEX_DIR, RSEM_GENE_INDEX);
    final String outputPrefix = String.format("%s.rsem", sampleId);
    // run RSEM
    StringBuilder rsemArgs = new StringBuilder();
    rsemArgs.append(" --alignments");
    rsemArgs.append(" --paired-end");
    rsemArgs.append(String.format(" %s", transcriptomeBam));
    rsemArgs.append(String.format(" %s", rsemGeneIndex));
    rsemArgs.append(String.format(" %s", outputPrefix));
    rsemArgs.append(String.format(" -p %s", threadCount));
    // run RSEM transcript expression calcs
    startupScript.addCommand(() -> format("%s/%s/%s %s", VmDirectories.TOOLS, RSEM_TOOL, RSEM_EXPRESSION_CMD, rsemArgs.toString()));
    startupScript.addCommand(() -> format("mv %s.rsem.genes.results %s.rsem.gene_data.tsv", sampleId, sampleId));
    startupScript.addCommand(() -> format("mv %s.rsem.isoforms.results %s.rsem.trans_data.tsv", sampleId, sampleId));
    startupScript.addCommand(new OutputUpload(GoogleStorageLocation.of(bucket.name(), "rsem"), executionFlags));
    // copy results to rna-analysis location on crunch
    startupScript.addCommand(() -> format("gsutil -m cp %s/*tsv %s/%s/rsem/", VmDirectories.OUTPUT, RNA_COHORT_LOCATION_V37, sampleId));
    return ImmutableVirtualMachineJobDefinition.builder().name("rna-rsem").startupCommand(startupScript).namespacedResults(ResultsDirectory.defaultDirectory()).workingDiskSpaceGb(500).performanceProfile(VirtualMachinePerformanceProfile.custom(12, 36)).build();
}

Example 28 with InputFileDescriptor

use of com.hartwig.batch.input.InputFileDescriptor in project pipeline5 by hartwigmedical.

the class SageCreatePonData method execute.

@Override
public VirtualMachineJobDefinition execute(final InputBundle inputs, final RuntimeBucket runtimeBucket, final BashStartupScript startupScript, final RuntimeFiles executionFlags) {
    final ResourceFiles resourceFiles = ResourceFilesFactory.buildResourceFiles(RefGenomeVersion.V37);
    final InputFileDescriptor remoteReferenceFile = inputs.get("reference");
    final InputFileDescriptor remoteReferenceIndex = remoteReferenceFile.index();
    final String localReferenceFile = localFilename(remoteReferenceFile);
    final String localReferenceBam = localReferenceFile.replace("cram", "bam");
    final String referenceSampleName = inputs.get("referenceSample").inputValue();
    // Download latest jar file
    // startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s %s",
    // "gs://batch-sage-validation/resources/sage.jar",
    // "/opt/tools/sage/" + Versions.SAGE + "/sage.jar"));
    // Download normal
    startupScript.addCommand(() -> remoteReferenceFile.toCommandForm(localReferenceFile));
    startupScript.addCommand(() -> remoteReferenceIndex.toCommandForm(localFilename(remoteReferenceIndex)));
    final SageCommandBuilder sageCommandBuilder = new SageCommandBuilder(resourceFiles).ponMode(referenceSampleName, localReferenceBam);
    final SageApplication sageApplication = new SageApplication(sageCommandBuilder);
    // Convert to bam if necessary
    if (!localReferenceFile.equals(localReferenceBam)) {
        startupScript.addCommands(cramToBam(localReferenceFile));
    }
    // Run post processing (NONE for germline)
    final SubStageInputOutput postProcessing = sageApplication.apply(SubStageInputOutput.empty(referenceSampleName));
    startupScript.addCommands(postProcessing.bash());
    // Store output
    startupScript.addCommand(new OutputUpload(GoogleStorageLocation.of(runtimeBucket.name(), "sage"), executionFlags));
    return VirtualMachineJobDefinition.sageSomaticCalling(startupScript, ResultsDirectory.defaultDirectory());
}

Example 29 with InputFileDescriptor

use of com.hartwig.batch.input.InputFileDescriptor in project pipeline5 by hartwigmedical.

the class SagePanelTumor method execute.

@Override
public VirtualMachineJobDefinition execute(final InputBundle inputs, final RuntimeBucket runtimeBucket, final BashStartupScript startupScript, final RuntimeFiles executionFlags) {
    InputFileDescriptor descriptor = inputs.get();
    final String sampleId = descriptor.inputValue();
    final ResourceFiles resourceFiles = ResourceFilesFactory.buildResourceFiles(RefGenomeVersion.V38);
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_TOOLS_BUCKET, SAGE_DIR, SAGE_JAR, VmDirectories.TOOLS));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_RESOURCE_BUCKET, SAGE_DIR, PANEL_BED, VmDirectories.INPUT));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_TOOLS_BUCKET, PAVE_DIR, PAVE_JAR, VmDirectories.TOOLS));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/38/* %s", BATCH_RESOURCE_BUCKET, GNOMAD_DIR, VmDirectories.INPUT));
    String ponFile = "SageGermlinePon.98x.38.tsv.gz";
    String ponArtefactFile = "pon_panel_artefact.38.tsv";
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_RESOURCE_BUCKET, SAGE_DIR, ponFile, VmDirectories.INPUT));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_RESOURCE_BUCKET, SAGE_DIR, ponArtefactFile, VmDirectories.INPUT));
    // download tumor BAM
    final String tumorBam = String.format("%s.non_umi_dedup.bam", sampleId);
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s* %s", PANEL_BAM_BUCKET, tumorBam, VmDirectories.INPUT));
    final String sageVcf = String.format("%s/%s.sage.somatic.vcf.gz", VmDirectories.OUTPUT, sampleId);
    // run Sage
    final StringJoiner sageArgs = new StringJoiner(" ");
    sageArgs.add(String.format("-tumor %s", sampleId));
    sageArgs.add(String.format("-tumor_bam %s/%s", VmDirectories.INPUT, tumorBam));
    sageArgs.add(String.format("-hotspots %s", resourceFiles.sageSomaticHotspots()));
    sageArgs.add(String.format("-panel_bed %s", resourceFiles.sageSomaticCodingPanel()));
    sageArgs.add(String.format("-high_confidence_bed %s", resourceFiles.giabHighConfidenceBed()));
    sageArgs.add(String.format("-ref_genome %s", resourceFiles.refGenomeFile()));
    sageArgs.add(String.format("-ref_genome_version %s", resourceFiles.version().toString()));
    sageArgs.add(String.format("-ensembl_data_dir %s", resourceFiles.ensemblDataCache()));
    sageArgs.add(String.format("-coverage_bed %s/%s", VmDirectories.INPUT, PANEL_BED));
    sageArgs.add(String.format("-out %s", sageVcf));
    sageArgs.add(String.format("-hotspot_min_tumor_qual 100"));
    sageArgs.add(String.format("-panel_min_tumor_qual 200"));
    sageArgs.add(String.format("-high_confidence_min_tumor_qual 200"));
    sageArgs.add(String.format("-low_confidence_min_tumor_qual 300"));
    sageArgs.add(String.format("-mnv_filter_enabled false"));
    sageArgs.add(String.format("-perf_warn_time 50"));
    // sageArgs.add(String.format("-log_debug"));
    sageArgs.add(String.format("-threads %s", Bash.allCpus()));
    startupScript.addCommand(() -> format("java -Xmx48G -jar %s/%s %s", VmDirectories.TOOLS, SAGE_JAR, sageArgs.toString()));
    // annotate with Pave - PON, Gnomad and gene impacts
    /*
        String bcfTools = String.format("%s/bcftools/%s/bcftools", VmDirectories.TOOLS, Versions.BCF_TOOLS);
        String ponVcf = String.format("%s/%s.sage.somatic.pon.vcf.gz", VmDirectories.OUTPUT, sampleId);

        // /data/tools/bcftools/1.9/bcftools annotate -a /data/resources/bucket/sage/37/SageGermlinePon.1000x.37.vcf.gz
        // -c PON_COUNT,PON_MAX
        // FR16648814.sage.somatic.vcf.gz
        // -O z
        // -o FR16648814.sage.somatic.annotated.vcf.gz

        final StringJoiner ponArgs = new StringJoiner(" ");
        ponArgs.add(String.format("-a %s", resourceFiles.sageGermlinePon()));
        ponArgs.add("-c PON_COUNT,PON_MAX");
        ponArgs.add(String.format("%s", sageVcf));
        ponArgs.add("-O z");
        ponArgs.add(String.format("-o %s", ponVcf));

        startupScript.addCommand(() -> format("%s annotate %s", bcfTools, ponArgs.toString()));

        ///data/tools/bcftools/1.9/bcftools filter
        // -e ‘PON_COUNT!=“.” && INFO/TIER=“HOTSPOT” && PON_MAX>=5 && PON_COUNT >= 5’
        // -s PON -m+ FR16648814.sage.somatic.annotated.vcf.gz -O u
        // | /data/tools/bcftools/1.9/bcftools filter -e ‘PON_COUNT!=“.” && INFO/TIER=“PANEL” && PON_MAX>=5 && PON_COUNT >= 2’
        // -s PON -m+ -O u | /data/tools/bcftools/1.9/bcftools filter -e ‘PON_COUNT!=“.” && INFO/TIER!=“HOTSPOT” && INFO/TIER!=“PANEL” && PON_COUNT >= 2’
        // -s PON -m+ -O z -o FR16648814.sage.somatic.pon_filtered.vcf.gz

        String ponFilterVcf = String.format("%s/%s.sage.somatic.pon_filter.vcf.gz", VmDirectories.OUTPUT, sampleId);

        // private static final String HOTSPOT = "INFO/TIER=\"HOTSPOT\" && PON_MAX>=%s && PON_COUNT >= %s";
        // private static final String PANEL = "INFO/TIER=\"PANEL\" && PON_MAX>=%s && PON_COUNT >= %s";
        // private static final String OTHER = "INFO/TIER!=\"HOTSPOT\" && INFO/TIER!=\"PANEL\" && PON_COUNT >= %s";

        final StringJoiner ponFilterArgs = new StringJoiner(" ");
        ponFilterArgs.add("-e 'PON_COUNT!=\".\" && INFO/TIER=\"HOTSPOT\" && PON_MAX>=5 && PON_COUNT >= 5'");
        ponFilterArgs.add(String.format("-s PON -m+ %s -O u", ponVcf));
        ponFilterArgs.add(String.format("| %s filter -e 'PON_COUNT!=\".\" && INFO/TIER=\"PANEL\" && PON_MAX>=5 && PON_COUNT >= 2'", bcfTools));
        ponFilterArgs.add(String.format("-s PON -m+ -O u | %s filter -e 'PON_COUNT!=\".\" && INFO/TIER!=\"HOTSPOT\" && INFO/TIER!=\"PANEL\" && PON_COUNT >= 2'", bcfTools));
        ponFilterArgs.add(String.format("-s PON -m+ -O z -o %s", ponFilterVcf));

        startupScript.addCommand(() -> format("%s filter %s", bcfTools, ponFilterArgs.toString()));
        */
    // finally run Pave
    final StringJoiner paveArgs = new StringJoiner(" ");
    String ponFilters = "HOTSPOT:5:5;PANEL:2:5;UNKNOWN:2:0";
    final String paveVcf = String.format("%s/%s.sage.somatic.pon.pave_pass.vcf.gz", VmDirectories.OUTPUT, sampleId);
    paveArgs.add(String.format("-sample %s", sampleId));
    // ponFilterVcf from BCF Tools
    paveArgs.add(String.format("-vcf_file %s", sageVcf));
    paveArgs.add(String.format("-ref_genome %s", resourceFiles.refGenomeFile()));
    paveArgs.add(String.format("-ref_genome_version %s", resourceFiles.version().toString()));
    paveArgs.add(String.format("-driver_gene_panel %s", resourceFiles.driverGenePanel()));
    paveArgs.add(String.format("-ensembl_data_dir %s", resourceFiles.ensemblDataCache()));
    paveArgs.add("-only_canonical");
    paveArgs.add("-filter_pass");
    paveArgs.add(String.format("-gnomad_freq_dir %s", VmDirectories.INPUT));
    paveArgs.add(String.format("-pon_file %s/%s", VmDirectories.INPUT, ponFile));
    paveArgs.add(String.format("-pon_artefact_file %s/%s", VmDirectories.INPUT, ponArtefactFile));
    paveArgs.add(String.format("-pon_filters \"%s\"", ponFilters));
    paveArgs.add("-gnomad_load_chr_on_demand");
    paveArgs.add(String.format("-output_vcf_file %s", paveVcf));
    String paveJar = String.format("%s/%s", VmDirectories.TOOLS, PAVE_JAR);
    // String paveJar = String.format("%s/pave/%s/pave.jar", VmDirectories.TOOLS, Versions.PAVE);
    startupScript.addCommand(() -> format("java -jar %s %s", paveJar, paveArgs.toString()));
    // upload output
    startupScript.addCommand(new OutputUpload(GoogleStorageLocation.of(runtimeBucket.name(), "sage"), executionFlags));
    return ImmutableVirtualMachineJobDefinition.builder().name("sage").startupCommand(startupScript).performanceProfile(custom(24, 64)).namespacedResults(ResultsDirectory.defaultDirectory()).build();
}

Example 30 with InputFileDescriptor

use of com.hartwig.batch.input.InputFileDescriptor in project pipeline5 by hartwigmedical.

the class SageRerun method execute.

@Override
public VirtualMachineJobDefinition execute(final InputBundle inputs, final RuntimeBucket runtimeBucket, final BashStartupScript startupScript, final RuntimeFiles executionFlags) {
    InputFileDescriptor descriptor = inputs.get();
    final String sampleId = descriptor.inputValue();
    final ResourceFiles resourceFiles = ResourceFilesFactory.buildResourceFiles(RefGenomeVersion.V37);
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_TOOLS_BUCKET, SAGE_DIR, SAGE_JAR, VmDirectories.TOOLS));
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_TOOLS_BUCKET, PAVE_DIR, PAVE_JAR, VmDirectories.TOOLS));
    String ponFile = "SageGermlinePon.1000x.37.tsv.gz";
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp %s/%s/%s %s", BATCH_RESOURCE_BUCKET, SAGE_DIR, ponFile, VmDirectories.INPUT));
    // download tumor and ref CRAM
    final RemoteLocationsApi locations = new RemoteLocationsApi("hmf-crunch", sampleId);
    String[] tumorCramData = getCramFileData(locations.getTumorAlignment());
    String tumorCramFile = tumorCramData[CRAM_FILENAME];
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp gs://%s* %s", tumorCramData[CRAM_FULL_PATH], VmDirectories.INPUT));
    String referenceId = locations.getReference();
    String[] refCramData = getCramFileData(locations.getReferenceAlignment());
    String refCramFile = refCramData[CRAM_FILENAME];
    startupScript.addCommand(() -> format("gsutil -u hmf-crunch cp gs://%s* %s", refCramData[CRAM_FULL_PATH], VmDirectories.INPUT));
    final String sageVcf = String.format("%s/%s.sage.somatic.vcf.gz", VmDirectories.OUTPUT, sampleId);
    // run Sage
    final StringJoiner sageArgs = new StringJoiner(" ");
    sageArgs.add(String.format("-tumor %s", sampleId));
    sageArgs.add(String.format("-tumor_bam %s/%s", VmDirectories.INPUT, tumorCramFile));
    sageArgs.add(String.format("-reference %s", referenceId));
    sageArgs.add(String.format("-reference_bam %s/%s", VmDirectories.INPUT, refCramFile));
    sageArgs.add(String.format("-hotspots %s", resourceFiles.sageSomaticHotspots()));
    sageArgs.add(String.format("-panel_bed %s", resourceFiles.sageSomaticCodingPanel()));
    sageArgs.add(String.format("-high_confidence_bed %s", resourceFiles.giabHighConfidenceBed()));
    sageArgs.add(String.format("-ref_genome %s", resourceFiles.refGenomeFile()));
    sageArgs.add(String.format("-ref_genome_version %s", resourceFiles.version().toString()));
    sageArgs.add(String.format("-ensembl_data_dir %s", resourceFiles.ensemblDataCache()));
    sageArgs.add(String.format("-out %s", sageVcf));
    sageArgs.add(String.format("-perf_warn_time 50"));
    // sageArgs.add(String.format("-log_debug"));
    sageArgs.add(String.format("-threads %s", Bash.allCpus()));
    startupScript.addCommand(() -> format("java -Xmx48G -jar %s/%s %s", VmDirectories.TOOLS, SAGE_JAR, sageArgs.toString()));
    // annotate with Pave - PON and gene impacts
    final StringJoiner paveArgs = new StringJoiner(" ");
    String ponFilters = "HOTSPOT:5:5;PANEL:2:5;UNKNOWN:2:0";
    final String paveVcf = String.format("%s/%s.sage.somatic.pon.pave.vcf.gz", VmDirectories.OUTPUT, sampleId);
    paveArgs.add(String.format("-sample %s", sampleId));
    // ponFilterVcf from BCF Tools
    paveArgs.add(String.format("-vcf_file %s", sageVcf));
    paveArgs.add(String.format("-ref_genome %s", resourceFiles.refGenomeFile()));
    paveArgs.add(String.format("-ref_genome_version %s", resourceFiles.version().toString()));
    paveArgs.add(String.format("-driver_gene_panel %s", resourceFiles.driverGenePanel()));
    paveArgs.add(String.format("-ensembl_data_dir %s", resourceFiles.ensemblDataCache()));
    paveArgs.add(String.format("-pon_file %s/%s", VmDirectories.INPUT, ponFile));
    paveArgs.add(String.format("-pon_filters \"%s\"", ponFilters));
    paveArgs.add(String.format("-output_vcf_file %s", paveVcf));
    String paveJar = String.format("%s/%s", VmDirectories.TOOLS, PAVE_JAR);
    startupScript.addCommand(() -> format("java -jar %s %s", paveJar, paveArgs.toString()));
    // upload output
    startupScript.addCommand(new OutputUpload(GoogleStorageLocation.of(runtimeBucket.name(), "sage"), executionFlags));
    return ImmutableVirtualMachineJobDefinition.builder().name("sage").startupCommand(startupScript).performanceProfile(custom(24, 64)).namespacedResults(ResultsDirectory.defaultDirectory()).build();
}