Microhaplotype discovery and genotyping

Goal

Discover microhaplotype block boundaries — either by LD-based haploblock detection (Jonas et al. 2017) or by fixed-window segmentation — and emit the final microhaplotype genotype matrix. This matrix is consumed downstream by masreml::build_G_mh() and masbayes::construct_wah_matrix().

TipPipeline position
  • Inputs: per-chromosome haploid files hap/chr* and the SNP map produced by the previous stage.
  • Outputs: block definition files and a numerical microhaplotype genotype matrix (one row per individual, two columns per block).
  • Tool: haplotype-hybrid (a maspipeline binary).
  • Previous stage: Genotype-to-haplotype
  • Next stage: — (pipeline output; ready for masreml / masbayes). See Input data → Microhaplotypes for the on-disk schema consumed by the R packages.

Because our objective is to discover microhaplotype segments, we only use outputs from hap/chr* and perform arguments --method ld-haploblock and --haplotype-type micro. The haplotype-hybrid tool works with finding all potential haplotype candidates that meet an LD criteria (e.g., \(D' > 0.45\)) as the argument of --d-prime-threshold 0.45 is applied. After that, align with the microhaplotype definition, they are selected from the haplotype block candidates, where in each segment should contain consecutive SNPs within 125 or 150 bp.

The final genotype data will be:

ID       hap_1_1  hap_1_1  hap_1_2  hap_1_2  hap_1_3  hap_1_3
ind_1        2        1        1        2        3        3
ind_2        1        3        2        2        3        2
ind_3        3        1        2        2        2        3

This genotype data consists of numerical microhaplotype alleles in diploid format, where each number represents a unique combination of SNP sequences in a specific genomic block. Each locus is presented in two side-by-side columns to show the pair of haplotypes inherited from each individual’s parents.

For example, at locus hap_1_1, individual ind_1 has alleles 2 and 1, meaning they inherited microhaplotype variant 2 from one parent and variant 1 from the other, where each variant represents a distinct multi-SNP sequence pattern (e.g., variant 1 might be “AACG” while variant 2 is “ATCG” for a 4-SNP microhaplotype).

It is important to note that, as we apply multiple criteria, generated microhaplotype segments will be significantly lower than fully haplotype blocks. The main advantage of this system is that we effectively use lower computational resources than haplotype-based genotypes or even biallelic SNPs. Therefore, a specialised coding matrix and model are strongly required to handle such genotype data, particularly to prevent multicollinearity while effectively utilising the multiallelic power.

Usage Example

# Microhaplotype discovery and genotyping (LD-based, physical window)
./haplotype-hybrid \
      --ncores 4 \
      -i hap/chr* \
      -m map.txt \
      -o mh_info_ld_micro \
      --generate-genotypes mh_genotypes \
      -v \
      ld-haploblock micro --window-bp 125

# Fixed physical distance blocks (e.g., 100 kb)
./haplotype-hybrid \
      -i hap/chr* \
      -m map.txt \
      -o mh_info_fixedkb \
      --generate-genotypes fixedkb_genotypes \
      -v \
      fixed-kb --window-bp 100000

Key Parameters

Use ./haplotype-hybrid --help to see manuals.

  • -i, --input <INPUT>...: Phased haplotype files, one per chromosome (required, e.g., hap/chr*)
  • -m, --map <MAP>: SNP map file with columns: SNPID, Chr, Position (required)
  • --method <METHOD>: Block definition method (default: ld_haploblock, possible values: ld-haploblock, snp-count-simple)
  • --haplotype-type <HAPLOTYPE_TYPE>: For LD method - pure selects best N-SNP haplotype per block, micro splits blocks into physical windows (default: pure, possible values: pure, micro)
  • --d-prime-threshold <D_PRIME_THRESHOLD>: D’ threshold for consecutive SNP LD in haploblock discovery (default: 0.45, Jonas 2017)
  • --window-bp <WINDOW_BP>: Physical window size in base pairs for microhaplotype segments (default: 125)
  • -w, --window <WINDOW>: Haplotype size in SNPs for pure mode or window size for snp-count-simple (default: 4)
  • --min-snps <MIN_SNPS>: Minimum SNPs required per block (default: 2)
  • --max-snps <MAX_SNPS>: Maximum SNPs allowed per block (default: 4)
  • --aft <AFT>: Allele Frequency Threshold for Criterion-B score calculation (default: 0.08)
  • --md <MD>: Maximum Deviation parameter for Criterion-B weighting (default: 0.1)
  • --min-ld <MIN_LD>: Minimum mean LD (r²) threshold for block filtering (optional, e.g., 0.3)
  • --no-dedup: Skip deduplication of blocks with identical SNP ranges
  • --noheader: Specify if haplotype input files have no header row
  • -o, --output <OUTPUT>: Output directory for block definition files (default: hap_info_microhap)
  • --generate-genotypes <GENERATE_GENOTYPES>: Generate haplotype genotype files with specified prefix (e.g., mh_geno_ld)
  • --missing <MISSING>: Code for missing alleles in genotype output (e.g., 9)
  • --missing-value <MISSING_VALUE>: Replacement value for blocks containing missing alleles (default: -9999)
  • -v, --verbose: Display detailed progress and diagnostic information
  • -h, --help: Print help information

What’s next?

← Back to overview.

The microhaplotype genotype matrix produced here is the input expected by masreml::build_G_mh() (additive GRM construction) and masbayes::construct_wah_matrix() (Wαh design matrix for BayesA / BayesR). See Input data → Microhaplotypes for the on-disk schema.