colockit

colockit is an R package for genomic colocalisation analysis with a clear two-stage workflow:

1. prepare_regions() — harmonises GWAS summary statistics within a locus, aligns effect alleles, and computes an LD matrix from a PLINK reference panel.
2. colocalise() — runs the chosen colocalisation method on the prepared data.

Supported methods:

Method	Description
"abf"	coloc Approximate Bayes Factor
"susie"	coloc SuSiE (multiple causal variants)
"pwcoco"	PWCoCo — external command-line tool (Robinson et al. 2022)

Supported backends:

Backend	Description
"standalone"	Supply GWAS summary statistics directly — no database required
"genoscores"	Integrate with the Genoscores database for published/private GWAS

Installation

remotes::install_github("molepi-precmed/colockit")

Or from a local checkout:

remotes::install_local("path/to/colockit")

Quick start

Standalone backend

library(colockit)
library(data.table)

locus <- data.table(
  chrom    = 22L,
  startpos = 29001000L,
  endpos   = 29011000L
)

## gwas1 and gwas2 are data.tables with columns:
## snp, dbsnpid, chrom, pos, allele, oth.allele,
## beta, pvalue, n_samples, freq, trait.name
gwas1 <- fread("trait1_summary_stats.csv")
gwas2 <- fread("trait2_summary_stats.csv")

## Stage 1 — harmonise and compute LD
regions <- prepare_regions(
  locus        = locus,
  gwas1        = gwas1,
  gwas2        = gwas2,
  output_dir   = "results/locus_22_29Mb",
  refplinkfile = "path/to/kg.2020.hg38.eur",
  backend      = "standalone"
)

## Stage 2 — colocalise
result <- colocalise(
  regions       = regions,
  locus         = locus,
  analysis_type = "abf",
  output_dir    = "results/locus_22_29Mb"
)

result$summary

Genoscores backend

con <- DBI::dbConnect(...)   # Genoscores database connection

regions <- prepare_regions(
  locus        = locus,
  gwas1        = 42L,         # integer gwasid
  gwas2        = 57L,
  output_dir   = "results/locus_22_29Mb",
  refplinkfile = "path/to/kg.2020.hg38.eur",
  backend      = "genoscores",
  gwas_type    = "published",
  con          = con
)

Private GWAS can be created with create_private_gwas() and used with gwas_type = "mixed" or gwas_type = "private".

Output files

prepare_regions() writes to output_dir:

• gwas1.csv, gwas2.csv — harmonised summary statistics
• ld.matrix.Rdata.gz — LD correlation matrix
• locus.plot.png — preview locus plot

colocalise() additionally writes:

• formatted.region.Rdata.gz — formatted coloc input
• coloc.abf.png / coloc.susie.png — colocalisation plot coloured by PP.H4

Development

Tests and R CMD check are run manually on a server with the 1000 Genomes reference panel and PWCoCo installed:

./run-tests.sh            # pull, restore deps, run tests + R CMD check
./run-tests.sh --no-pull  # skip git pull (use current working tree)

Learn more

• vignette("getting-started") — full walkthrough of both flows
• vignette("simulated-colocalisation") — reproducible example using coloc::coloc_test_data with known causal structure (PP.H4 ≈ 1)

References

Robinson JW, Hemani G, Babaei MS, Huang Y, Baird DA, Tsai EA, Chen C-Y, Gaunt TR, Zheng J (2022). An efficient and robust tool for colocalisation: Pair-wise Conditional and Colocalisation (PWCoCo). bioRxiv. https://doi.org/10.1101/2022.08.08.503158

License

MIT © 2026 Andrii Iakovliev