Skip to content

DGRPool v1.0.0 — DGRP3 call set, platform upgrade and data releases

Latest

Choose a tag to compare

@fabdavid fabdavid released this 20 Nov 14:26
· 5 commits to main since this release

Highlights

  • Curation complete. Every paper in DGRPool has now been reviewed by a curator, so no study is listed as "under curation" anymore.
  • DGRP3 genotypes added (dm6, 416 lines). DGRPool now uses the combined DGRP2+DGRP3 call set: 4,724,310 variants × 416 genotyped lines, annotated with SnpEff 5.4 / FlyBase r6.64. Variant pages and searches are now dm6-only. User GWAS jobs submitted earlier against DGRP2/dm3 have been re-queued on the new call set.
  • Versioned data releases. A new Releases section lists each data release with its date, its downloadable files (DGRP lines, phenotypes, studies, all phenotype means by sex) and the GWAS command used.
  • Better on phones and small screens. The navigation bar, header, footer, home page and phenotype table now adapt to small screens.

GWAS

  • User GWAS now uses the DGRP3 genotypes and covariates with MAF 0.05. MAF and missingness are computed only on the lines that have a phenotype. PLINK2 options now match the precomputed GWAS: BETA is reported for the ALT allele, and -9 is treated as a real value.
  • Logistic top_* threshold corrected to 0.001.
  • PheWAS loading skips variants that are no longer present and removes outdated results.
  • New index on gwas_results.snp_id makes variant pages and gene pages faster.

DGRP lines

  • A line's status now only says whether it is genotyped. Available lines are no longer counted as genotyped.
  • Each genotype is now mapped to its line through a stored position (dgrp_lines.geno_index), so editing a line's status can no longer mismatch genotypes and lines.
  • Lines can no longer be created or edited from the web interface.

Studies, phenotypes and datasets

  • Publication year now comes from the publication date (published_at), falling back to issued when it is missing.
  • Fixed encoding of author names and initials.
  • Authors and titles edited by hand are no longer overwritten.
  • studies.tsv export now includes volume, issue, journal name and all authors.
  • Only admins can delete datasets.
  • Fixed:
    • re-uploading a dataset with fewer lines could remove line links from other studies
    • deleting a summary dataset left its phenotypes behind
    • raw re-uploads marked every existing phenotype as "bad"
    • cached data was not cleared after a delete
  • Fixed sex_by_dgrp, plus a backfill task for existing data.
  • Statistics are recomputed when a category is added, and when a phenotype's name or numeric flag changes (not on every update).
  • Statistics are computed daily by cron and saved to one file per day. A day's file is skipped if nothing changed.

Platform and security

  • Upgraded to Ruby 4.0.9, Rails 8.1.4 and R 4.6.0.
    • All 138 R packages now have fixed versions (CRAN / Bioconductor 3.23).
    • PLINK2 v2.0.0-b1.4, with checksum verification.
    • Gemfile.lock is now committed.
  • Fixed CSS/JS not loading after the Rails 8 upgrade: compression is now handled by Rack::Deflater.
  • Security fixes, including:
    • path traversal blocked in file downloads
    • secure session cookie
    • dependency vulnerability updates
    • admin and curator emails no longer stored in the code
  • Fixes to the check-your-phenotype tool, GWAS analysis pages, the JSON endpoints used by download_phenotypes.R, and the Solr reindex.