How to run colocalization pipeline

Describe how to run the new colocalization pipeline with coloc susie package

Introduction

This pipeline takes the outputs from our finemapping pipeline, and perform colocalization among 571 resources we gathered, including all GWAS endpoints from FinnGen, UKB, eQTL catelogue, Generisk project, proteomics study from INTERVAL, UKB and FinnGen.

Data Source	Data type	Description
FinnGen-R12	GWAS	all endpoints from FinnGen R12
GeneRisk	GWAS	GeneRISK Study is an ongoing prospective observational study focusing on genetic risk factors of cardiovascular diseases and on utilizing genetic information in preventing diseases.
UKB-finucane	GWAS	Some endpoints from UKB shared from Masahiro. https://www.medrxiv.org/content/10.1101/2021.09.03.21262975v1
Alasoo_2018--macrophage_naive--ge	eQTL_Catalogue	expression QTL from eQTL catalogue (release 6), gathered from macrophage and based on gene expression, see eQTL catelogue website for more information
... (other ~560 more items)	eQTL_Catelogue	Other resources from eQTL Catelogue indicated by the data source. eQTL catelogue assembled multiple data sources, e.g., tissue expression from GTEX.
INTERVAL	Plasma-Proteomics	Proteomics QTL from INTERVAL
UKB-PPP	Plasma-Proteomics	Proteomics QTL from UKBiobank (Olink)
FIN-R12-Olink	Plasma-Proteomics	Proteomics QTL from FinnGen R12 (Olink)
FIN-R12-Somascan	Plasma-Proteomics	Proteomics QTL from FinnGen R12 (Somascan)

Example to run

1. Download the meta data from finemapping pipeline.

Menu(Applications) -> Sandbox -> pipelines and find your successful finemapping run -> click download metadata (assumed to be located in Downloads/XXXX_metadata.json)

2. Submit the colocalization job in local terminal in the sandbox

# run the script: metadata, trait_name, data_type, storage bucket (your green bucket)
# please customize those inputs to your own project and data_type can be any string wihout space)
# please change the red bucket number "N" to match your sandbox environment, you can see the red bucket uri by running "gsutil ls" in SB terminal 
/finngen/shared_nfs/finngen/coloc/submit ~/Downloads/XXXX_metadata.json T2D GWAS gs://fg-production-sandbox-"N"-red/YOUR_PATH/T2D_Project

Check the errors if there are some.

If no error occurs, pressing the Enter key at the terminal will open a browser to check the jobs. Refresh and look into your submitted job. The job is named "ColocSusieDirectMulti" with your user name, it takes some time to show due to reponse time for the backends in the sandbox.

3. Download results

The outputs are labeled as "ColocSusieDirectMulti.colocQC" in output of pipeline's job details. We only keep the H4.PP > 0.5 and valid credible set from both dataset (the threshold could be controled in the input). Future filtering should be performed based on your purpose to this output, e.g., H4.PP > 0.8 and overlapped region size. We could not provide a gold standard for this, as it is dependent on the study design and the aim for colocalization.

The raw results are listed in the "ColocSusieDirectMulti.coloc" without any filtering and merging.

"ColocSusieDirectMulti.hit": all the information for the top signals in the full colocalization results.

"ColocSusieDirectMulti.pairs": the overlapped region being run in the workflow.

Output formats

Column	Description
dataset1	generated from your trait_name and data_type
dataset2	Study--DataType in our resources
trait1	the trait name in your data
trait2	trait name / molecular phenotype name from our resources
region1	region in your data
region2	overlapped region in our resources
cs1	credible set in your data
cs2	credible set in our resources
nsnps	total variants overlapped
hit1	top signal in your data
hit2	top signal in our resources
PP.H4.abf	probability of colocalization between your data and our resources
low_purity1	the credible set is low purity or not in your data. (1 means low purity, 0, high purity)
low_purity2	the purity in our resources
nsnps1	number of variants in region from your data
nsnps2	number of variants in region from our resources
cs1_log10bf	log10 bayes factor for the credible set in your data
cs2_log10bf	log10 bayes factor for the credible set in our resources
clpp	colocalization based on CLPP
clpa	colocalization based on CLPA (min of PIP)
cs1_size	size of the raw credible set in your data
cs2_size	size of the raw credible set in our resources
cs_overlap	size of the overlapped credible set
topInOverlap	Indicator if a top variant (highest PIP) in each dataset is in the overlap region of finemapped regions of the 2 datasets. 1,1: both orginal top signal located in the overlapped region (expected reasonable coloc); 1,0 /0,1: only one top in the overlapped region; 0,0: both top signal are not in the overlapped.
hit1_info	information of top signal in your data (beta, p-value)
hit2_info	information of top signal in our resources (beta, p-value)

Codes are available on github: https://github.com/FINNGEN/coloc.susie.direct