Processed FASTQ
Paired-end processed FASTQ files (*.fastp.fastq.gz) are accompanied by fastp QC reports in JSON and HTML formats. Salmon uses only these processed FASTQ files; transcriptome BAM is not used for quantification.
Analysis outputs
Output roles and verification status
The output manifest and workflow records show whether each required file was generated and checked. The presence of a filename alone is not considered evidence that an analysis completed successfully.
Preprocessing and alignment
BAM / BAI
The coordinate-sorted BAM and its index are checked with quickcheck and for consistency of the header, reference sequences, and alignment records.
STAR outputs
Log.final.out, SJ.out.tab, ReadsPerGene.out.tab, and the transcriptome BAM are stored with the software and version information used to generate them.
Counts, abundance estimates, and expression matrices
FeatureCounts
GENCODE 49 biotype-level QC uses exon and gene_type. This output complements rather than replaces STAR GeneCounts.
Salmon
For every selected profile, Harako-GPU records quant.sf, quant.genes.sf, runtime, version, container image, index, options, and library type.
Expression matrices and concordance
Outputs include transcript- and gene-level expression matrices for TPM, estimated reads (NumReads), and effective length. compare-both also reports abundance differences and rank concordance between the two Salmon versions.
Reports and execution records
| Output | Purpose |
|---|---|
| MultiQC | Summarizes the defined QC results and provides access to the underlying metrics. |
| Nextflow report, timeline, trace, and DAG | Record workflow execution, but do not by themselves prove that the complete analysis finished successfully. |
backend-output-manifest.json | Records each output's role, sample, relative path, size, and SHA-256 checksum independently of the selected backend. |
| Output manifest | Marks required outputs as generated, verified, or not applicable. |
| Support information | Collects the logs and state needed for troubleshooting while excluding FASTQ, BAM, expression matrices, and credentials. |
To protect your data, do not send or publish FASTQ, BAM, expression matrices, reference data, indices, complete trace logs, credentials, or unedited support material in a public issue or public AI service.
Storing completed runs
Active output files and the Nextflow work directory must be placed on fast local storage.
After successful completion and verification, results can be explicitly saved to high-capacity archive storage. Work from a failed or resumable run remains on local storage until the problem is resolved or the user explicitly abandons the run.
The complete Nextflow work directory may be removed only through a separate user action after the archive has been verified. Harako-GPU v0.1.0-alpha.1 does not archive or delete data automatically, and archive storage is not used as the Nextflow workDir.
Technical specification: harako-gpu-terminal-results-archive-v1
The archive record verifies the archive contents and checksums. It does not establish biological validity or determine whether the data may be published.