Zubair Khalid

Virologist/Molecular Biologist | Veterinarian | Bioinformatician

Conventional & Molecular Virology • Vaccine Development • Computational Biology

Dr. Zubair Khalid is a veterinarian and virologist specializing in conventional and molecular virology, vaccine development, and computational biology. Dedicated to advancing animal health through innovative research and multi-omics approaches.

Dr. Zubair Khalid - Veterinarian, Virologist, and Vaccine Development Researcher specializing in Computational Biology, Multi-omics, Animal Health, and Infectious Disease Research

Category: Guides

Phase Genomics: A Practical Guide to Haplotype Phasing

Phase genomics, also called haplotype phasing, is the computational process of assigning individual sequence variants to the maternal or paternal copy of each chromosome. This guide explains when and how to phase genomic data, what methods exist, and how to avoid common pitfalls. It is written for bioinformaticians, genomic researchers, and advanced graduate students who plan to incorporate phasing into their variant analysis workflows.

At a Glance

Aspect Detail
Core concept Determining which alleles reside on the same chromosome copy
Primary data Whole genome sequencing, linked reads, or long reads
Key methods Statistical phasing (e.g., Beagle, SHAPEIT), trio phasing, read based phasing
Recommended for Imputation, expression QTL studies, compound heterozygote detection, population genetics
Not for Most single variant association studies without downstream haplotype analysis
Main limitation Requires dense variant data or family structure, accuracy drops in low complexity regions

What Is Phase Genomics and Why Does It Matter

Genome sequencing generates a list of variants but initially loses the information of which variants originated from the same physical chromosome. Phase genomics recovers that information. For example, if a person carries two nearby missense variants, phasing can tell you whether they lie on the same copy (cis) or on different copies (trans). This distinction is critical for interpreting compound heterozygosity, allele specific expression, and meiotic recombination maps. The NCBI Bookshelf provides an accessible introduction to the fundamentals of genome variation and phasing in its genetic analysis chapters NCBI Bookshelf.

A phased genome also improves imputation accuracy. When you phase a study cohort, you can match haplotypes against a reference panel such as the 1000 Genomes Project. This step increases the number of testable variants in downstream analyses like GWAS. The EMBL EBI training materials offer practical exercises on haplotype estimation and imputation using real data EMBL EBI Training.

Decision Criteria: When to Phase

Not every project needs phasing. Consider phasing when any of the following apply:

  • You plan to impute genotypes: Phasing is a prerequisite for most imputation pipelines.
  • You study compound heterozygosity in recessive disorders: You need to know whether two rare variants are on opposite copies.
  • You perform allele specific expression analysis: Phase information anchors which transcript comes from which parental allele.
  • You work with population or family data: Phasing helps reconstruct ancestral haplotypes and detect selection signals.

Avoid phasing when you have only a sparse set of variants (e.g., from a targeted panel) and no reference panel, or when your samples are highly degraded. The Galaxy Training Network provides decision trees for choosing between statistical and read based phasing strategies based on input data type Galaxy Training Network.

Practical Phasing Workflow

A typical phasing workflow involves five stages.

1. Obtain High Quality Variant Calls

Phasing requires accurate variant positions and genotypes. Use a standard GATK or FreeBayes pipeline with hard filtering or VQSR. The NCBI Sequence Read Archive holds thousands of sequenced genomes that you can use as test data or for benchmarking NCBI Sequence Read Archive. Ensure your VCF file contains phased or unphased genotypes with proper allele depth annotations.

2. Determine Phasing Method

Three main approaches exist.

  • Statistical phasing: Uses population linkage disequilibrium to infer haplotypes. Suitable for large cohorts (hundreds to thousands). Tools include Beagle 5.4, SHAPEIT5, and Eagle2. These require a reference panel or a large enough study cohort to estimate haplotype frequencies.
  • Trio phasing: Uses parent child genotype data to assign phase deterministically. Works with any number of variants but requires family samples.
  • Read based phasing: Uses long reads, linked reads, or Hi C data to physically connect variants on the same molecule. Accurate but more expensive and computationally intensive.

For most projects, start with statistical phasing using a tool like Beagle. The Bioconductor project includes the package rbio for converting VCF files into formats compatible with phasing software Bioconductor.

3. Run the Phasing Software

Example command for Beagle (version 5.4):

java -jar beagle.22Jul22.46e.jar
  gt=unphased.vcf.gz
  ref=reference.panel.vcf.gz
  map=plink.chr1.GRCh38.map
  out=phased_chr1

Adjust memory and threads as needed. For SHAPEIT5 use the shapeit5_phase_common or shapeit5_phase_rare subcommands depending on variant frequency.

4. Validate Phasing Quality

Check the output VCF file. Most phasing tools report a dosage or probability for each phased call. Use the following measures.

  • Switch error rate: Compare phased calls against known trio data. An error below 2 percent is typical for good quality whole genome data.
  • Mean sample haplotype length: Very short haplotypes may indicate poor phasing.
  • Concordance with long read data: If available, validate a subset of phased positions.

5. Use Phased Data

Once phased, you can proceed to imputation, haplotype based association tests, or visualization with tools like IGV or the R package haploR. Store phased VCF with the PS (phase set) tag for downstream compatibility.

Quality Checks

Always run the following checks before trusting phased output:

  • Verify that the input VCF contains only biallelic SNPs and indels. Multiallelic sites confuse many phasing algorithms.
  • Check for strand flips. If your reference panel uses a different genome build, liftover your data first.
  • Assess sample relatedness. Statistical phasing assumes unrelated individuals. Including close relatives can bias haplotype estimates.

The Galaxy Training Network offers tutorials on quality control metrics for VCF files that apply directly to phasing inputs Galaxy Training Network.

Common Mistakes

Several mistakes recur in phasing projects.

  • Using too small a cohort: Statistical phasing requires at least 50 to 100 samples to estimate linkage disequilibrium accurately.
  • Ignoring reference panel compatibility: The panel must match your data in genome build, variant representation, and population origin.
  • Mixing sequencing platforms: Combining short and long read data without careful filtering can introduce systematic errors.
  • Forgetting to phase X and Y chromosomes separately: Most tools treat autosomes and sex chromosomes with different parameters.
  • Overinterpreting phase around structural variants: Breakpoints may not be covered by standard SNP phasing tools.

Limits and Uncertainty

No phasing method is perfect. Statistical phasing has difficulty in regions with low recombination or recent admixture. Read based phasing requires high coverage and does not fill gaps between distant variants. Trio phasing is the gold standard but is often unavailable. Always report the phasing method and error rate in publications. The EMBL EBI training materials emphasize that phase accuracy decreases for rare variants and in repetitive regions EMBL EBI Training.

Phase information can also change with updated reference panels. A dataset phased with an older panel may need re phasing when a new panel is released. Treat phased haplotypes as probabilistic estimates, not deterministic truths.

Frequently Asked Questions

Q: Can I phase a single genome without a reference panel or family data?
A: No. Statistical phasing requires population data. Read based phasing with long reads can phase a single genome but only for variants that lie on the same sequencing read or read cloud. Most single genome projects therefore remain unphased for rare variants.

Q: What is the minimum coverage required for read based phasing?
A: At least 30x for short reads with linked read technology, or 15x for long reads. Lower coverage reduces the chance that two variants appear on the same read.

Q: How do I choose between Beagle and SHAPEIT?
A: Beagle is simpler and handles large datasets well. SHAPEIT5 offers better performance for rare variants and includes a specific module for very rare variants. Both produce similar results on common variants.

Q: Does phasing improve GWAS power?
A: Indirectly. Phased data enable haplotype based association tests that can capture effects of multiple causal variants within a gene. For single variant tests, phasing does not directly increase power, but it improves imputation which can add new markers.

References and Further Reading

Related Articles