polyqtlR (version 0.1.1)

impute_dosages: Re-estimate marker dosages given IBD input estimated using a high error prior.

Description

Function to correct marker dosage scores given a list of previously estimated IBD probabilities. This may prove useful to correct genotyping errors. Running the estimate_IBD function with a high error prior will result in suppressed predictions of double recombination events, associated with genotyping errors. By forcing the HMM to penalise double recombinations heavily, a smoothed haplotype landscape is achieved in which individual genotype observations are down-weighted. This smoothed output is then used to re-estimate marker dosages, dependent on (correct) parental scores. An alternative strategy is to use the function maxL_IBD over a range of error priors first, and use the resulting $maxL_IBD output as input here (as the IBD_list). In this case, set the argument min_error_prior to a low value (0.005 say) to avoid issues.

Usage

impute_dosages(
  IBD_list,
  dosage_matrix,
  parent1 = "P1",
  parent2 = "P2",
  rounding_error = 0.05,
  min_error_prior = 0.1,
  verbose = TRUE
)

Arguments

IBD_list

List of IBD probabilities

dosage_matrix

An integer matrix with markers in rows and individuals in columns. Note that probabilistic genotypes are not currently catered for here.

parent1

The identifier of parent 1, by default "P1"

parent2

The identifier of parent 2, by default "P2"

rounding_error

The maximum deviation from an integer value that an inputed value can have, by default 0.05. For example, an imputed score of 2.97 or 3.01 would both be rounded to a dosage of 3, while 2.87 would be deemed too far from an integer score, and would be made missing. If you find the output contains too many missing values, a possibility would be to increase the rounding_error. However this may also introduce more errors in the output!

min_error_prior

Suggestion for a suitably high error prior to be used in IBD calculations to ensure IBD smoothing is achieved. If IBD probabilities were estimated with a smaller error prior, the function aborts.

verbose

Should messages be written to standard output?

Examples

if (FALSE) {
# Toy example only, as this will result in an Error: the original error prior was too low
data("IBD_4x","SNP_dosages.4x")
impute_dosages(IBD_list=IBD_4x,dosage_matrix=SNP_dosages.4x)
}