Phasing

Goal

Resolve diploid genotypes (e.g. 0/1) into haploid maternal/paternal strands (e.g. 0|1 and 1|0) so microhaplotype blocks can later be read off each strand directly. Phasing is done per chromosome in parallel with Beagle; bcftools splits the genome and tabix indexes the per-chromosome outputs.

TipPipeline position

• Inputs: cleaned, biallelic, unphased VCF (single file, or already split per chromosome).
• Outputs: a phased VCF (genotypes_phased.vcf) with genotypes encoded as 0|0, 0|1, 1|0, 1|1.
• Tool: Beagle (external; we tested with beagle.22Jul22.46e.jar), bcftools, tabix. No maspipeline binary required for this stage — the pipeline calls Beagle directly.
• Previous stage: Data preparation
• Next stage: Genotype-to-haplotype

---

Phasing is the first critical step in our pipeline, where ambiguous diploid genotypes are separated into their constituent haploid sequences to allow for the accurate identification of microhaplotype blocks. We encourage to do phasing using Beagle in parallel process, and to do that, we use bcftools to split the VCF file per chromosome and tabix for indexing a compressed VCF, as an example:

for chr in {1..29}; do
    bcftools view -r ${chr} /data/genotypes.vcf.gz -O z -o /data/tmp/chr${chr}.vcf.gz
    tabix -p vcf /data/tmp/chr${chr}.vcf.gz
done

for chr in {1..29}; do
    (
        java -Xmx28g -jar ${EBROOTBEAGLE}/beagle.jar \
            gt=/data/tmp/chr${chr}.vcf.gz \
            nthreads=4 \
            out=/data/tmp/chr${chr}_phased
    ) &

    if (( chr % 8 == 0 )); then
        wait
    fi
done
wait

for chr in {1..29}; do
    tabix -p vcf /data/tmp/chr${chr}_phased.vcf.gz
    bcftools reheader -h <(bcftools view -h /data/genotypes.vcf.gz) /data/tmp/chr${chr}_phased.vcf.gz | \
        bcftools view -O z -o /data/tmp/chr${chr}_phased_corrected.vcf.gz
    tabix -p vcf /data/tmp/chr${chr}_phased_corrected.vcf.gz
done

bcftools concat -O z -o /data/genotypes_phased.vcf.gz \
    /data/tmp/chr{1..29}_phased_corrected.vcf.gz

tabix -p vcf /data/genotypes_phased.vcf.gz

/usr/bin/rm -rf /data/tmp
gunzip /data/genotypes_phased.vcf.gz

This process results in a phased VCF file, for example:

##fileformat=VCFv4.2
##filedate=20260108
##source="beagle.22Jul22.46e.jar"
##INFO=<ID=AF,Number=A,Type=Float,Description="Estimated ALT Allele Frequencies">
##INFO=<ID=DR2,Number=A,Type=Float,Description="Dosage R-Squared: estimated squared correlation between estimated REF dose [P(RA) + 2*P(RR)] and true REF dose">
##INFO=<ID=IMP,Number=0,Type=Flag,Description="Imputed marker">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##FORMAT=<ID=DS,Number=A,Type=Float,Description="estimated ALT dose [P(RA) + 2*P(AA)]">
#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  Ind_1   Ind_2
1   16977   snp1    C   A   .   .   PR  GT  0|0 1|1
1   33723   snp2    G   T   .   .   PR  GT  1|0 0|0

---

What’s next?

→ Continue to Genotype-to-haplotype — split each phased genotype into its two haploid columns with convert-from-vcf.

← Back to overview.