vignettes/IntroductionCoPheScan_01.Rmd
IntroductionCoPheScan_01.RmdThe cophescan package implements Coloc adapted Phenome-wide Scan (CoPheScan), a Bayesian method to perform Phenome-wide association studies (PheWAS) that identifies causal associations between genetic variants and phenotypes while simultaneously accounting for confounding due to linkage disequilibrium.
Given a query variant and genomic region with Q SNPs for a query trait, cophescan discriminates between 3 hypotheses:
\(H_n\) : No association with the query trait (1 configuration)
\(H_a\) : Association of a variant other than the query variant with the query trait (Q-1 configurations)
\(H_c\) : Association of the query variant with the query trait (1 configuration)
with \(p_n\), \(p_a\) and \(p_c\) being their corresponding priors.
cophescan can be run in different ways depending on the size and type of dataset.
First, choosing the method for Bayes factor computation:
| Single causal variant | Multiple causal variants | Requires LD matrix | |
|---|---|---|---|
| ABF | ✔ | x | No |
| SuSIE | ✔ | ✔ | Yes |
Whenever, LD matrices are available (preferably in-sample LD),
`cophe.susie` is the recommended method as it accounts for
multiple causal variants in the tested region.
Next, depending upon the size of the dataset we choose the method to specify priors :
| Dataset | Inclusion of covariates | |
|---|---|---|
| Fixed priors | Small | - |
| Hierarchical priors | Large | ✔ |
The different combinations that can be run are:
ABF/Fixed priors: cophe.single
SuSIE BF/Fixed priors: cophe.susie
ABF/Hierarchical priors: cophe.single.lbf +
run_metrop_priors
SuSIE BF/Hierarchical priors: cophe.susie.lbf +
run_metrop_priors
Description of the CoPheScan method:
CoPheScan: phenome-wide association studies accounting for linkage disequilibrium
coloc: Giambartolomei et al (2013)
coloc with SuSIE: Wallace et al (2021), github
ABF: Wakefield (2008)
SuSIE: Wang et al (2020), github