docs: add Claude Code context files and DRAGEN pipeline reference - #39
docs: add Claude Code context files and DRAGEN pipeline reference#39joshuamschmidt wants to merge 3 commits into
Conversation
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…/dragen_guide.md in CLAUDE.md
michael-harper
left a comment
michael-harper
left a comment
There was a problem hiding this comment.
This looks good and will definitely save tokens. I'm wondering if docs/dragen_guide.md is actually allowed to be distributed like this, is it proprietary?
| --qc-coverage-count-soft-clipped-bases true | ||
| --qc-coverage-reports-1 cov_report,cov_report | ||
| --qc-coverage-filters-1 'mapq<1,bq<0,mapq<1,bq<0' | ||
| --vc-gvcf-gq-bands 13 20 30 40 |
There was a problem hiding this comment.
These have just been changed to 10 20 30 40
| ``` | ||
| --qc-coverage-reports-1 cov_report,cov_report | ||
| --qc-coverage-filters-1 'mapq<1,bq<0,mapq<1,bq<0' | ||
| --vc-gvcf-gq-bands 13 20 30 40 |
There was a problem hiding this comment.
These have just been changed to 10 20 30 40
| ### QC metric CSVs — expected | ||
|
|
||
| MultiQC collects these via `rglob('*.csv')` — there is no validation step; missing files are silently | ||
| absent from the report rather than causing an error. |
There was a problem hiding this comment.
I believe MultiQC is going to be removed from this pipeline as it lives in the single_sample_qc_popgen pipeline
…t into analy and sis throughout the doc.
| | gVCF mode | `vc_emit_ref_confidence` | `GVCF` | gVCF only; excluded from bulk download | | ||
| | VCF output | `vc_enable_vcf_output` | `false` | No VCF, only gVCF | | ||
| | CNV calling | `enable_cnv` | `true` | | | ||
| | CNV segmentation | `cnv_segmentation_mode` | `SLM` | Shifting Level Models | |
There was a problem hiding this comment.
Just a note, in WGS this is SLM, in WES it is recommended to set to HSLM (Heterogeneous SLM). How would we like to capture these kinds of differences?
| pytest test | ||
|
|
||
| # Bump version (updates Dockerfile, config TOML, README, creates git tag) | ||
| bump2version <major|minor|patch> |
There was a problem hiding this comment.
This might differ among systems. I use bump-myversionbecause it looked likebumpversionandbump2versnio` were both deprecated when I set it up originally.
3 participants
Summary
CLAUDE.md— Claude Code project instructions giving the AI assistant full context on architecture, the stage DAG, key modules, configuration, and how to navigate the DRAGEN 3.7 guide. Includes a section map of guide line numbers for topics relevant to this pipeline (QC outputs, caller options, input/reference, full option reference).docs/dragen_guide.md— the full DRAGEN 3.7 user guide (21 103 lines), checked in so Claude Code can Read specific sections by offset without hitting the filesystem outside the repo.docs/dragen_pipeline_reference.md— a concise, repo-maintained reference for our specific pipeline configuration: enabled modules, mode-specific inputs, expected output files (QC CSVs, variant/call files, what's excluded from bulk download), ICA parameter conventions, and config key reference.docs/dragen_pipeline_reference.mdalso lists ICA help doc URLs in .md format that Claude Code can fetch directly via WebFetch for live ICA platform documentation (API reference, Nextflow pipeline details, CLI data transfer, etc.) without needing a checked-in copy.Why
This helps those who use our DRAGEN outputs and potential contributors (and Claude) understand what DRAGEN outputs to expect, what gets downloaded vs skipped, and how ICA parameters map to DRAGEN CLI flags.
Previously this lived only in the 21k-line DRAGEN guide and in scattered code comments. This surfaces the relevant subset in a single place and keeps it versioned with the code.