SCALLOP-INF meta-analysis¶

A companion web site for this paper¹,

Zhao, J.H., Stacey, D., Eriksson, N., Macdonald-Dunlop, E., Hedman, Å.K., Kalnapenkis, A., Enroth, S., Cozzetto, D., Digby-Bell, J., Marten, J., Folkersen, L., Herder, C., Jonsson, L., Bergen, S.E., Gieger, C., Needham, E.J., Surendran, P., Metspalu, A., Milani, L., Mägi, R., Nelis, M., Hudjašov, G., Paul, D.S., Polasek, O., Thorand, B., Grallert, H., Roden, M., Võsa, U., Esko, T., Hayward, C., Johansson, Å., Gyllensten, U., Powell, N., Hansson, O., Mattsson-Carlgren, N., Joshi, P.K., Danesh, J., Padyukov, L., Klareskog, L., Landén, M., Wilson, J.F., Siegbahn, A., Wallentin, L., Mälarstig, A., Butterworth, A.S., Peters, J.E., and Estonian Biobank Research Team (2023). Genetics of circulating inflammatory proteins identifies drivers of immune-mediated disease risk and therapeutic targets. Nature Immunology, URL https://www.nature.com/articles/s41590-023-01588-w.

Quick links to codes for figures,

Flow of analysis¶

The diagram can also be rendered via Mermaid live editor.

Comments¶

To view the code inside the browser, select the GitHub button from the menu.

The tryggve, cardio and csd3 directories here are associated with the named Linux cluster(s) used for the analysis over time. Early implementation involves the following aspects,

1. Data pre-processing from tryggve with list.sh and format.sh, followed by meta-analysis according to metal.sh using METAL whose results were cross-examined with QCGWAS.sh together with additional investigation.
2. The main analysis with analysis.sh containing codes for Manhattan/Q-Q/forest/LocusZoom plots, clumping using PLINK and conditional analysis using GCTA. The clumping results were classified into cis/trans signals. As the meta-analysis stabilised especially with INTERVAL reference, analysis has been intensively done locally with cardio and csd3. cis/trans classification has been done via cis.vs.trans.classification.R as validated by cistrans.sh.
3. Prototyping analysis on cardio with INTERVAL such as INTERVAL.sh and cardio.sh as well as individual level data analysis for the KORA study. Most analyses were done locally on CSD3.

Most recent implementations are documented in Supplementary notes (rsid). Over time, many functions become part of two R packages, gap (CRAN, GitHub, vignette) and pQTLtools (Web page).

A benchmark¶

As revealed by PhenoScanner², Osteoprotegerin (OPG) proves to be a positive control³ involving both a cis and a trans pQTLs, with the cis-pQTL showing stronger association -- see a stacked image containg forest+LocusZoom and Manhattan+Q-Q plots, OPG.pdf.

Summary statistics¶

They will be available from

• Cardiovascular Epidemiology Unit (CEU), https://www.phpc.cam.ac.uk/ceu/proteins
• GWAS catalog, https://www.ebi.ac.uk/gwas/publications/37563310 (accession GCST90274758-GCST90274848 [gcst_list.xlsx], protein-target-gene mapping)
  • https://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90274001-GCST90275000/
  • https://www.ebi.ac.uk/gwas/studies/GCST90274758

    …

    https://www.ebi.ac.uk/gwas/studies/GCST90274848

The diagram is based on circos in the named directory circos which highlights pQTLs [causal genes]; the significance levels of association can be seen from the inner scatter plot whose ceiling for -log10(P) is set to be 150.

Table: PAVs in +/-1MB regions & r2>=0.8¶

References¶

.