Genotype to haplotype

Goal

Bridge between phased VCF and the per-chromosome haploid CSV format that haplotype-hybrid expects. Two maspipeline binaries handle the round trip: convert-to-vcf (CSV → VCF, used during Data preparation if you start from CSV), and convert-from-vcf (phased VCF → per-chromosome haploid files). This page documents convert-from-vcf — the post-phasing step that this stage represents.

TipPipeline position
  • Inputs: phased VCF (output of Phasing).
  • Outputs: per-chromosome haploid files (hap/chr*.hap), per-chr genotype files (geno/chr*.geno), and an SNP map.
  • Tool: convert-from-vcf (a maspipeline binary).
  • Previous stage: Phasing
  • Next stage: Microhaplotype discovery

After phasing, we want to separate the phased haplotypes into individual haploid sequences for haploblock detection. Here, we provide convert-from-vcf for conversion, which accomplishes two key transformations:

1. Haplotype separation

Each phased genotype (e.g., 0|1) is split into two separate haplotype columns:

  • Individual_A with genotype 0|1 becomes:
    • Haplotype 1: 1 (recoded from 0)
    • Haplotype 2: 2 (recoded from 1)

This separation is essential because LD-based haploblock detection requires calculating pairwise \(D'\) between SNPs, which depends on counting the some possible haplotype combinations (e.g. 0|0, 0|1, 1|0, 1|1).

2. Allele recoding

VCF format uses 0 (reference) and 1 (alternate) encoding. The conversion recodes these to:

  • 0 → 1 (major allele)
  • 1 → 2 (minor allele)

For genotypes (diploid coding):

  • 0|0 → 0 (homozygous major)
  • 0|1 or 1|0 → 1 (heterozygous)
  • 1|1 → 2 (homozygous minor)

To handle large datasets, individuals are processed in chunks (batches) rather than all at once. The chunk size is determined by the --interval parameter:

  • --interval 1: Process all individuals in one pass (high memory)
  • --interval 10: Process in 10 batches (lower memory, slightly slower)

Each chunk reads through the VCF file sequentially, extracts the relevant individuals, and writes output incrementally. Data is automatically split by chromosome, creating separate output files (haplotypes and genotypes), for example:

Haplotype Files (hap/chr*.hap)

Format: Each individual contributes two columns (one per haplotype)

ID                snp1_h1  snp1_h2  snp2_h1  snp2_h2  ...
Ind_1             1        2        1        1        ...
Ind_2             2        2        1        2        ...

This provides input for LD-based haploblock detection.

Genotype Files (geno/chr*.geno)

Format: Traditional genotype matrix with header

ID              snp1  snp2  snp3  ...
Ind_1           1     0     2     ...
Ind_2           2     1     1     ...

This provides input for GBLUP or other genotype-based prediction models that utilises biallelic SNPs.

Usage Example

./convert-from-vcf \
      -i /data/genotypes_phased.vcf.gz \
      --map map.txt \
      --genofolder geno \
      --hapfolder hap

Key Parameters

  • -i, --input: VCF file(s) to convert (required)
  • --interval: Number of chunks for processing (default: 1)
  • --hapfolder: Output directory for haplotype files (default: hap)
  • --genofolder: Output directory for genotype files (default: geno)
  • --map: Output path for SNP map file (default: map_new.txt)
  • --missing: Code for missing genotypes (default: -9999)
  • -v, --verbose: Display detailed progress information
  • -h, --help: Print help information

The separated haplotype files (hap/chr*) then serve as direct input for the next step: defining microhaplotype segments using LD-based haploblock detection.

What’s next?

→ Continue to Microhaplotype discovery — discover microhaplotype block boundaries and emit the final genotype matrix.

← Back to overview.