coloc.abf: Fully Bayesian colocalisation analysis using Bayes Factors

Description

Bayesian colocalisation analysis

Usage

coloc.abf(dataset1, dataset2, MAF = NULL, p1 = 1e-04, p2 = 1e-04, p12 = 1e-05)

Arguments

dataset1

a list with the following elements

pvalues: P-values for each SNP in dataset 1
N: Number of samples in dataset 1
MAF: minor allele frequency of the variants
beta: regression coefficient for each SNP from dataset 1
varbeta: variance of beta
type: the type of data in dataset 1 - either "quant" or "cc" to denote quantitative or case-control
s: the proportion of samples in dataset 1 that are cases (only relevant for case control samples)
snp: a character vector of snp ids, optional. If present, it will be used to merge dataset1 and dataset2. Otherwise, the function assumes dataset1 and dataset2 contain results for the same SNPs in the same order.

Some of these items may be missing, but you must give type and then either pvalues, N and s (if type="cc") or beta and varbeta. If you use pvalues, then the function needs to know minor allele frequencies, and will either use the MAF given here or a global estimate of MAF supplied separately.

dataset2

as above, for dataset 2

MAF

Common minor allele frequency vector to be used for both dataset1 and dataset2

prior probability a SNP is associated with trait 1, default 1e-4

prior probability a SNP is associated with trait 2, default 1e-4

p12

prior probability a SNP is associated with both traits, default 1e-5

Value

a list of two data.frames:

summary is a vector giving the number of SNPs analysed, and the posterior probabilities of H0 (no causal variant), H1 (causal variant for trait 1 only), H2 (causal variant for trait 2 only), H3 (two distinct causal variants) and H4 (one common causal variant)
results is an annotated version of the input data containing log Approximate Bayes Factors and intermediate calculations, and the posterior probability SNP.PP.H4 of the SNP being causal for the shared signal

Details

This function calculates posterior probabilities of different causal variant configurations under the assumption of a single causal variant for each trait.

If regression coefficients and variances are available, it calculates Bayes factors for association at each SNP. If only p values are available, it uses an approximation that depends on the SNP's MAF and ignores any uncertainty in imputation. Regression coefficients should be used if available.