Colocalization of two proteins

[dariushghasemi]

Here we perform colocalization analysis using the coloc v5.2.3 (2023-09-22).
We follow some steps here:

• we cross tabulate credible set SNPs for two independent SNPs for two different proteins,
• once there is s common SNP between the two credible sets, we store corresponding conditional data of the two independent SNPs.
• repeat this analysis for all combinations of independent signals from one proteins against all other proteins to attain all possible combinations of the traits we need to perform coloc,
• conduct main colocalization analysis of two traits using the input created in previous steps.

[dariushghasemi]

The coloc analysis with previous version of the Nicola's pipe has been run successfully. Here we try to update the function computing l-ABF to the latest version removing NULL from the conditional dataset.

[dariushghasemi]

We combined coloc info tables using R script workflow/scripts/s06_collect_info.R rather than in bash. The scripts:

• scans all the files (RDS/TSV/sentinel etc.) created by rule run_cojo stored in results/*/cojo/
• takes info tables (TSV)
• for the input seqids (extracted from input sentinel files), selects info table for input seqids from all present COJO outputs
• reads and merges coloc info tables
• extracts chromosomal position from file names which helps iterate rule run_coloc in parallel for each chromosome.

The input/output/parameter directives of rule master_file. are changed accordingly.

[dariushghasemi]

In the next commit, we allocated resources in SLURM configuration slurm/config.yaml for each of the three rules running colocalization analysis in Snakemake.

[dariushghasemi]

In the following commit, we ran colocalization analysis on all proteins sequences with the current parameters in config file eg. hole=3Mb. Some statistics about the results of this preliminary version of coloc pipeline are:

• size: 119 Mb
• nrows: 593,014
• ncols: 12
• n. of unique loci: 3367
• n. of unique seqids: 3720

---

We save the pipe output in this directory:
/exchange/healthds/pQTL/results/META_CHRIS_INTERVAL/qced_sumstats_digits_not_flipped_filtered/14-Jan-25_combined_colocalization_results.csv

Along with the coloc results, a READMe is also provided to briefly describe method, output column names, and config parameters at the above directory.

Column	Description
nsnps	number of variants shared between the two input loci
PP.H0.abf	posterior probabality for hypothesis 0: No causal variant
PP.H1.abf	posterior probabality for hypothesis 1: causal variant for t1
PP.H2.abf	posterior probabality for hypothesis 2: causal variant for t2
PP.H3.abf	posterior probabality for hypothesis 3: distinct causal variants for t1 and t2
PP.H4.abf	posterior probabality for hypothesis 4: shared causal variant for t1 and t2
t1	trait name of protein_1 (seqid)
t2	trait name of protein_2 (seqid)
target1	the most significant variant at locus_1
target2	the most significant variant at locus_2
locus1	locus belonging to protein_1
locus2	locus belonging to protein_2

---

NOTE: This analysis results are preliminary and may change later.