Library state bundle (2026-05-04)

Prebuilt library.sqlite and library_summary.json used by the Docker image. Generated from the upstream metadata CSVs by backend.library.ingest. The Docker workflow downloads these assets at build time so the container ships ready to demo without needing the 467 MB Lib4_info_concise source CSV. To refresh: re-run ingest locally, then gh release upload library-data-2026-05-04 --clobber library.sqlite library_summary.json.

TREND v0.1.0

Changelog

v0.1.0 — 2026-04-23

Initial release. Three artifacts in one repository:

pipeline/ — trend-pipeline CLI

• trend init — scaffold a project from ovarian_cancer or T_cell_activation templates
• trend run — execute the 9-step pipeline locally or via Snakemake on SLURM (--profile slurm)
• trend run --example {smoke,step9,pipeline} — three reviewer-facing reproducibility tiers
• trend dashboard — launch the web UI pointed at any runs directory
• trend status — one-screen run summary
• trend preflight — environment check with per-OS install hints
• Bundled Snakemake workflow with conda env definitions and SLURM profile
• Bioconda recipe at pipeline/conda-recipe/meta.yaml ready for submission

dashboard/ — interactive web dashboard

• Library composition view faithful to manuscript Figure 1 (panels A, B, C, D, E, F)
  • Library target composition pyramid (1,068 → 729 / 91 / 248 decomposition)
  • DBD family composition bar chart (49 Lambert families; 28 named bars + Other)
  • Sensors per DBD family bar chart
  • TREND coverage of CaCTS cancer master TFs across 34 TCGA tumor types
  • TREND coverage of D'Alessio cell identity TFs across 15 anatomical systems
  • Cancer-selectivity scatter for the OvCa project
• Enhancer table with sortable columns, free-text search, Lambert taxonomy badges, and click-to-filter from any panel
• Pipeline runner with 9-step state machine visualization
• Three-tier reproducibility check with green/red oracle badges
• Published-results browser with column-by-column tooltips
• System status page with FR-2 environment preflight
• Lovable-inspired warm cream / charcoal visual system (shadcn/ui + Tailwind)

tools/

• build_fixtures.py — deterministic fixture generator for the three reviewer tiers (subsamples real OvCa data + simulates FASTQs with planted activity profiles)

Reproducibility tiers

• Tier 1 — Quick check (1 s, no external tools): comparator validates against bundled published outputs
• Tier 2 — Activity reproduction (~30 s, requires R + tidyverse + Rsamtools): unchanged Step 9 R script reproduces published activity from a 1,000-promoter slice of real OvCa data
• Tier 3 — Full pipeline (~3 min, requires conda env): Steps 1-9 end-to-end on simulated FASTQs (50 promoters x 5 barcodes x 8 samples) with analytically-correct expected count matrix

Tests

• 18 passing tests under tests/ (pytest tests/ -v)
  • 7 csv-comparator unit tests (the C2 equivalence predicate)
  • 2 oracle E2E tests (OvCa + T-cell)
  • 9 trend CLI tests

Documentation

• README.md — short overview, install summary
• MANUAL.md — comprehensive user + reviewer manual (install, three-tier verification, dashboard tour, adopter walkthrough, troubleshooting, glossary)
• DASHBOARD_PRD.md — product requirements
• DESIGN.md — visual system specification
• references/TREND_library_TF_breakdown.md — TF composition reconciliation against Lambert / Reddy / D'Alessio

Data hosting

• Code + bundled fixtures: GitHub
• Full data (~3 GB: published alignment count tables + Lib4 reference + per-construct metadata): Dropbox, fetched by scripts/download_data.{sh,ps1}
• Post-acceptance: planned migration of full data to Zenodo with citable DOI