coloc.fast: Calculate colocalisation probabilities.
In tobyjohnson/gtx: Genetics ToolboX

Description Usage Arguments Details Value Author(s)

Calculates posterior probabilities for colocalisation of two association signals, using the approximate Bayes factor method of Giambartolomei et al.

1	coloc.fast(data, rounded = 6, priorsd1 = 1, priorsd2 = 1, priorc1 = 1e-4, priorc2 = 1e-4, priorc12 = 1e-5, beta1 = 'beta_1', se1 = 'se_1', beta2 = 'beta_2', se2 = 'se_2')

`data`	Data frame of association results.
`rounded`	Number of digits to round posterior probabilities, or NA for no rounding.
`priorsd1`	Standard deviation of the prior on true effect sizes for phenotype 1.
`priorsd2`	Standard deviation of the prior on true effect sizes for phenotype 2.
`priorc1`	Prior on a given variant being causal for phenotype 1.
`priorc2`	Prior on a given variant being causal for phenotype 2.
`priorc12`	Prior on a given variant being causal for phenotypes 1 and 2.
`beta1`	Name of effect size for phenotype 1 in data.
`se1`	Name of standard error for phenotype 1 in data.
`beta2`	Name of effect size for phenotype 2 in data.
`se2`	Name of standard error for phenotype 2 in data.

To be described. Equivalent to Giambartolomei method but uses a single framework for quantitative and binary phenotypes, and allows user control of the within-model priors on effect sizes (to which the results can be surprisingly sensitive).

The Giambartolomei package coloc uses a prior SD of 0.15 for quantitative traits, which cannot easily be changed by the user. This function coloc.fast, if called with non-default options priorsd1=0.15 and priorsd2=0.15, produces numerically identical results to coloc for every example I have checked.

alpha12 and alpha21 are estimates respectively of beta1/beta2 and of beta2/beta1, if the true effect sizes were known for the single causal variant under hypothesis 4.

A list with four elements: results is a data frame of posterior model probabilities; nvariants which reports the number of variants used in the analysis; alpha12 and alpha21 are estimates of the causal effect of each trait on the other (see below).

Toby Johnson Toby.x.Johnson@gsk.com

tobyjohnson/gtx documentation built on Aug. 30, 2019, 8:07 p.m.