v1.2.0 - CRITICAL FIX

General description

This release introduces a critical fix that addresses a bug in liftOver procedure applied to the BED/BIM/FAM input (see #110)

CRITICAL FIX

During the liftover of plink bfiles in s01_alpha_sort_alleles_snpid.R, we perform a merge between the old and new .bim files (to add the lifted positions), but this merge changes the order of the SNPs. Since we do not update the BED genotype matrix, this results in genotype mismatches and incorrect LD and allele frequencies.

This release introduces a crucial fix to maintain the original SNP order during the liftOver of BED dataset.

Note that overall results were not affected if you don't apply liftOver to the BED/BIM/FAM input.

v1.1.0

This release addresses a few bugs and introduces a couple of enhancements adding the possibility to select specific chromosomes for analysis and allowing better control over finemapping

Bug fixes

• fix a bug when both summary stats and bim file are in GRCh37 build and no liftover is requested (#95)
• correctly report as finemapping exceptions loci that are removed by finemapping QC (#91)
• update flandeRs R package to version 1.0.1, to avoid an error when shared credible-sets are found during the colocalization step. Now the pipeline will produce an empty guide table, and colocalization is skipped as expected.

Enchancements

• user can now select which chromosome to consider in the analysis using the chromosomes parameter (#86)
• refactor finemapping QC logic post SUSIE (see below)
• sdY can now be passed as a file reporting sdY for each trait (#85). This is useful when working with molecular traits, such as eQTLs

Refactored finemapping QC and SUSIE logic

1. Selective QC by L

Before: All loci (fine-mapped with either L=1 or L=10) were passed through post-SuSiE QC.

After:

• Only loci fine-mapped with L=10 undergo QC.
• Loci fine-mapped with L=1 are kept as-is (QC skipped).

2. Fallback strategy when QC removes all credible sets

Before: If all credible sets were removed by QC, the locus was silently dropped.

After:

• If QC removes all credible sets at L=10, the locus is re-fine-mapped with L=1 (and accordingly, QC is then skipped).
• Loci re-finemmaped at L=1 are reported in the finemapping_exceptions/FINEMAPPED_L1_recover_after_susie_QC.tsv file, stating "Locus re-finemapped at L=1 after none of the credible sets fine-mapped at L=10 passed post susie QC."

3. Switched from natural log to raw BF threshold

Before: BF threshold was passed as natural log Bayes factor and then converted to log₁₀ scale (log10(exp(cs_bf_thr)))
After: BF threshold is now passed as raw Bayes factor and then converted to log₁₀ scale (log10(cs_bf_thr))

v1.0.0 - First official release

First official release of Flanders pipeline ( Finemapping coLocalization AND plEiotRopy Solver).

The pipeline allows for large-scale colocalization analysis directly from summary statistics by:

• parallelized summary statistics harmonization and significant loci identification
• parallelized fine-mapping of significant loci using SUSIE
• efficient storing of a large-scale credible set in a dedicated data structure based on anndata
• efficient colocalization across a large-scale volume of credible sets stored in anndata dataset