SEPIA · document index

All documents

Everything as HTML — the plan, the data & pipeline record, the rebuild, and the per-genome evidence. Source of truth remains the Markdown in the repo; these are generated renderings (except the two hand-built pages).

Start here

Roadmap & plan →
The single phased plan: bugs, analysis redesign, QC (Q1–Q10), infra, writing. Hand-built.
sepia-plan.html · from plan.md
Data & pipeline (living) →
BAM vs VCF, the raw data, the data-flow, exact coverage, and the run log we update as we go. Hand-built.
pipeline.html
Current state & handoff →
Where we are right now: decisions locked in, what's in flight, open questions. Read this first on a cold pickup.
STATE.html · from STATE.md
Changelog & decision log →
Dated record of everything done / found / decided.
changelog.html · from changelog.md

Rebuild & execution

Rebuild from raw →
The clean-slate, QC-gated pipeline (Stages 0–8), reuse-vs-rerun matrix, compute estimates.
rebuild_from_raw.html
B2a runbook →
The fast build fix (panel→hg19): step-by-step commands, SLURM, compute/access needs.
b2a_runbook.html
Cluster access & Claude's rules →
How the SSH tunnel works, why it broke, robust config, and the binding rules Claude follows so it doesn't wedge it again.
CLUSTER_ACCESS.html

Which genomes, and why (provenance & QC)

Sample inclusion →
The justified subset — criteria + Poisson table; maximize reliable SNPs, not genome count.
papers/SAMPLE_INCLUSION.html
References & provenance →
Per-genome source paper, DOI, coverage, exact VCF/BAM/ENA URLs, data tier.
papers/REFERENCES.html
Per-paper QC & pipeline →
What each source actually did (verbatim params, adversarially verified).
papers/PIPELINE_QC_BY_PAPER.html
Published-VCF QC →
What QC each published VCF already carries + the harmonization we must add.
papers/VCF_QC.html

Evidence & audit

Bug evidence log →
Every bug reproduced + traced to its downstream consequence (demonstrated/none/latent).
verify/BUG_EVIDENCE.html
Pipeline audit workbook →
Step-by-step audit of the original pipeline (per-step I/O, commands, known issues). Lists cluster paths.
pipeline_workbook.html
Cluster inventory →
The 110 GB Great Lakes layout, the ~19-job pipeline, data products, access. Lists cluster paths.
cluster_inventory.html
Evaluation plan (retired) →
The QC items now tracked as Q1–Q10 in the roadmap; kept as a pointer.
evaluation_plan.html