compare_pangenome_covariates: compare_pangenome_covariates

In gtonkinhill/panstripe: Pangenome Plots

compare_pangenome_covariates

Description

Tests whether covariates associated with different pangenomes are significantly associated with gene gain and loss or errors.

Usage

compare_pangenome_covariates(
  fits,
  covariates,
  family = "Tweedie",
  keep = "all",
  modeldisp = FALSE,
  ci_type = "norm",
  conf = 0.95,
  nboot = 100
)

Arguments

fits	a list of 'panfit' objects generated by the 'panstripe' function
covariates	a 'data.frame' object generated where the first column matches the names in the list of pangenomes. Covariates to be tested are given in subsequent columns.
family	the family used by glm. One of 'Tweedie', 'Poisson', 'Gamma' or 'Gaussian'. (default='Tweedie')
keep	a vector of a subset of the column names in the covariate data.frame. (default='all')
modeldisp	whether or not to model the dispersion as a function of the covariates of interest if using a Tweedie family (default=FALSE)
ci_type	the method used to calculate the bootstrap CI (default='bca'). See boot.ci for more details.
conf	A scalar indicating the confidence level of the required intervals (default=0.95)
nboot	the number of bootstrap replicates to perform (default=100)

Value

a list containing a summary of the comparison and the resulting 'glm' model object

Examples

simA <- simulate_pan(rate=1e-4, ngenomes = 50, fn_error_rate=1, fp_error_rate=1)
simB <- simulate_pan(rate=1e-3, ngenomes = 200, fn_error_rate=1, fp_error_rate=1)
simC <- simulate_pan(rate=5e-3, ngenomes = 100, fn_error_rate=1, fp_error_rate=1)
tfits <- purrr::map(list(A=simA, B=simB, C=simC), ~{
  panstripe(.x$pa, .x$tree, nboot=10, ci_type='perc')
})

covariates <- tibble::tibble(
  pangenome=c('A','B','C','E','F','G'),
  dummy=c(1,2,3,1,2,2)
)
fits <- c(tfits, list(E=tfits[[1]], F=tfits[[2]], G=tfits[[3]]))
comp <- compare_pangenome_covariates(fits, covariates, modeldisp=TRUE)
comp$summary

gtonkinhill/panstripe documentation built on Feb. 27, 2025, 9:01 p.m.