COVID19 GWAS genes enrichment in disease-specific cells

[emdann]

Notes from chat with Shuang

Motivation:

do the disease-specific states (i.e. collection of neighbourhoods) show differential expression (or other type of association) with SNPs associated with disease? Can we use IDed disease-states for better characterization of GWAS hits compared to what we do on healthy cell types with fGWAS analysis?

GWAS summary-stats data

From the GWAS catalog (this only gives significant coefficients but you need all)

• https://www.ebi.ac.uk/gwas/publications/34237774
• https://www.ebi.ac.uk/gwas/publications/35255492

Methods

fGWAS approach

used by Natsuhiko/Patrick (code) (paper)

LD score regression

Papers

• Orig LD score (to check the presence of confounders in GWAS) https://www.nature.com/articles/ng.3211
• Partitioned LD score: partitioned by SNPs associated to different functional assays https://www.nature.com/articles/s41588-018-0081-4
• LD score analysis for scRNA-seq associations https://www.biorxiv.org/content/10.1101/2021.03.19.436212v1.full.pdf

Analysis examples: see LDSC package

• HLCA analysis (code)
• LD score computation “partitioned” for custom annotation
• Cell-type specific analysis (i.e. an example of partitioned LD score analysis) https://github.com/bulik/ldsc/wiki/Cell-type-specific-analyses

Notes:

• LD score is calculated by ethnicity, so it should be matched with ethnicity of cohort used for GWAS

[emdann]

Feedback from Mike: do this on another tissue, rather than PBMCs