S5 MagicImpute

MagicImpute

The magicimpute first imputes founders and then imputes offspring. During the iterative founder imputation, it simultaneously performs the following:

• Impute founder genotypes.
• Delete markers that do not fit in. By default, isdelmarker = true.
• Correct founder genotypes. By default, iscorrectfounder = true if model="depmodel" or offspring do not have genotypes in AD format`.
• Infer marker-specific error rates. By default, isinfererror = true if model is not "depmodel" or isspacemarker = true.
• Refine local marker ordering. By default, isordermarker = true if mapfile is not nothing.
• Refine inter-marker distances. By default, isspacemarker = true if mapfile is not nothing or isordermarker = true or isphysmap=true.

After founder imputation, magicimpute imputes offspring.

• If phasealg=forwardbackward, for each offspring at each marker, the posterior diplotype probabilities (in format GP) corresponding to the phased genotypes 0|0, 0|1, 1|0 and 1|1 are caculated according to the forward backward algorithm, and the called phased genotypes (in format GT) are given by those with the largest posterior diplotype probabilities if they are greater than threshimpute.
• If phaselag=viterbi(default), the posterior diplotype probabilities (GP) are set as those of phasealg=forwardbackward, and the called phased genotypes (GT) are calculated according to the Viterbi algorithm.
• If phaselag=unphase, the posterior genotype probabilities (GP) corresponding to the unphased genotypes 0/0, 0/1, and 1/1 are caculated by tranforming the posterior diplotype probabilities of phasealg=forwardbackward, and the called unphased genotypes are given by those with the largest posterior genotype probabilities if they are greater than threshimpute.

# code for Julia
using MagicImpute
cd(@__DIR__)
genofile = outstem*"_magicfilter_geno.vcf.gz"
pedfile = outstem*"_magicfilter_ped.csv"
magicmask_impute(genofile,pedfile;
    mapfile = outstem*"_magicmap_construct_map.csv.gz",          
    outstem     
)

# code for Linux shell. 
# For Window CMD, replace multiline key \ by  ^, and replace comment-key # by ::
julia rabbit_magicmask_impute.jl -g example_magicfilter_geno.vcf.gz \
    -p example_magicfilter_ped.csv \
    --mapfile example_magicmap_construct_map.csv.gz \
    --nworker 5 \
    -o example

Output files

Output: imputation accuracy

outstem*"_magicimpute_founderacc.csv" gives imputation accuracies for each founder, and outstem*"_magicimpute_offspringacc.csv" gives imputation accuracies for subpopulation, assuming that masked genotypes in outstem*"_magicmask_reversed.vcf.gz are ground truth. These output files ( + outstem*"_magicmask_geno.vcf.gz") do not exist for magicimpute.

using CSV, DataFrames
outstem = "example"
CSV.read(outstem*"_magicimpute_founderacc.csv",DataFrame; comment="##")

CSV.read(outstem*"_magicimpute_offspringacc.csv",DataFrame; comment="##")

Output: map refinement

outstem*"_magicimpute_map.csv" is the refined mapfile after refining local marker ordering and/or inter marker distance. outstem*"_magicimpute_compare_inputmap.png" compare the refined mapfile with the map of input genofile and the map of keyword mapfile (if no nothing). These output files do not exist if input map is not changed.