- Phasing and haplotype reconstruction are important processes in bioinformatics for determining how genetic variants are arranged on individual chromosomes. Most diploid organisms inherit one set of chromosomes from each parent. At a particular genomic region, an individual may carry different alleles on the two homologous chromosomes. Although conventional genotyping identifies the alleles present at each position, it does not always reveal which alleles occur together on the same chromosome. Phasing resolves this uncertainty by assigning variants to their corresponding parental or chromosomal copies, while haplotype reconstruction uses this information to infer the sequences of alleles carried along a chromosome or genomic region.
- A haplotype is a combination of alleles located on the same chromosome and inherited together as a unit over a particular genomic region. In a diploid individual, two homologous chromosomes may carry different haplotypes. For example, an individual may have two heterozygous variants at nearby positions, with alleles A/G at the first position and C/T at the second. The genotype data establish that the individual carries A and G at the first position and C and T at the second, but they do not necessarily establish how these alleles are arranged across the two chromosomes.
- There are two possible arrangements. In one arrangement, A and C occur together on one chromosome, while G and T occur together on the other. This is called the coupling phase or cis configuration. In the alternative arrangement, A and T occur together on one chromosome, while G and C occur together on the other. This is called the repulsion phase or trans configuration. Both arrangements produce the same unphased genotypes at the two positions, but they represent different haplotypes. Phasing determines which arrangement is supported by the available biological and sequencing evidence.
- The distinction between phased and unphased genotypes is represented in VCF files using different separators. A forward slash indicates an unphased genotype, while a vertical bar indicates a phased genotype. For example, 0/1 represents a heterozygous genotype whose alleles have not been assigned to specific chromosome copies, whereas 0|1 represents a phased genotype in which the allele ordering is meaningful relative to the chromosome copies represented in the file. Similarly, 1|0 and 0|1 describe different ordered allele assignments when the phase-set context is defined.
- A simplified VCF example illustrates the difference:
##fileformat=VCFv4.3
##FORMAT=<ID=GT,Number=1,Type=String,Description=”Genotype”>
#CHROM POS REF ALT QUAL FILTER INFO FORMAT Sample1
1 100 A G 60 PASS . GT 0/1
1 200 C T 60 PASS . GT 0/1
- The two genotypes are heterozygous, but their phase is unknown. The individual could carry the reference alleles together or carry one reference and one alternative allele on each chromosome. A phased representation might look like this:
##fileformat=VCFv4.3
##FORMAT=<ID=GT,Number=1,Type=String,Description=”Genotype”>
##FORMAT=<ID=PS,Number=1,Type=Integer,Description=”Phase set identifier”>
#CHROM POS REF ALT QUAL FILTER INFO FORMAT Sample1
1 100 A G 60 PASS . GT:PS 0|1:100
1 200 C T 60 PASS . GT:PS 0|1:100
- In this example, both variants have the same phase-set identifier, indicating that their allele ordering belongs to the same phase set. The first chromosome copy carries the reference alleles A and C, while the second carries the alternative alleles G and T. The exact biological identity of the two chromosome copies, such as maternal or paternal origin, is not established merely by using a vertical bar. Parental origin requires additional evidence, such as parental genotypes or other inheritance information.
- The VCF PS field is commonly used to identify a phase set. Variants with the same phase-set value are interpreted as belonging to the same phased block, within the relevant sample and chromosome context. A phase set indicates that the relative allele ordering is known within that block, but it does not necessarily establish the relationship between variants in different phase sets. Consequently, two separately phased blocks may have known internal haplotypes but an unknown relative orientation to each other. Other fields, including PGT and PID in some caller-specific outputs, may also contain phasing information, but their definitions and conventions should be checked in the file header and software documentation.
- Phasing can be performed using several types of biological and computational evidence. Read-based phasing uses sequencing reads that cover multiple heterozygous variants. If a single sequencing read contains two or more variant positions, the observed alleles provide evidence about which variants occur on the same DNA molecule. Paired-end reads can sometimes provide additional linkage information through their overlapping or paired alignment positions. Long-read sequencing is particularly useful because individual reads may span several variants across longer genomic distances, allowing larger haplotype blocks to be reconstructed.
- For example, suppose two heterozygous variants are located within the length of a sequencing read. If reads repeatedly contain the alternative allele at both positions, this supports their occurrence on the same chromosome copy. If reads instead show the alternative allele at the first position together with the reference allele at the second, the evidence may support a different phase configuration. Read-based phasing methods use these observations to construct groups of linked variants. The reliability of the inferred phase depends on read quality, sequencing depth, mapping accuracy, allele balance, and the number of informative reads spanning the variants.
- Short-read sequencing provides useful phasing information when variants are sufficiently close together to be covered by the same read or linked read pairs. However, typical short reads cannot directly connect variants separated by long genomic distances. As a result, short-read phasing may produce relatively small phase blocks. Repetitive regions, low sequencing coverage, mapping ambiguity, and sequencing errors can further reduce the reliability of read-based phase inference.
- Long-read sequencing technologies such as PacBio HiFi and Oxford Nanopore Technologies can improve haplotype reconstruction because their reads often span multiple heterozygous sites over longer distances. PacBio HiFi reads combine relatively long read lengths with high consensus accuracy, while Oxford Nanopore reads can provide very long molecules that connect distant variants. These technologies are useful for phasing complex genomic regions, resolving structural variants, reconstructing haplotypes, and studying genes with multiple nearby polymorphisms. Their performance still depends on coverage, read accuracy, genomic context, and the specific phasing method used.
- Trio-based phasing uses genetic information from an individual and both biological parents. By comparing the child’s genotypes with the parental genotypes, variants can often be assigned to the maternal or paternal chromosome. For example, if a child carries an alternative allele that is also present in the mother but absent in the father, the allele is generally assigned to the maternal chromosome, provided that the genotypes are reliable and inheritance is consistent with the assumed model. Trio information can therefore establish parental origin and connect variants across larger genomic regions than read-based phasing alone in some circumstances.
- Family-based phasing can also use information from siblings and other relatives. Pedigree relationships, Mendelian inheritance patterns, and recombination events can provide evidence about haplotype transmission. However, errors in pedigree information, genotyping errors, de novo mutations, recombination, and complex inheritance patterns may complicate the analysis. Trio-based phasing is particularly valuable in clinical genetics because it can help determine whether two pathogenic variants in the same gene are in cis or trans.
- Statistical phasing uses population-level patterns of genetic variation and linkage disequilibrium (LD) to infer the most likely haplotypes. Linkage disequilibrium describes the non-random association of alleles at different genomic positions within a population. If particular alleles frequently occur together in a population, statistical algorithms can use these patterns to estimate the phase of unphased genotypes. Statistical phasing is especially useful for large cohorts in which many individuals share common haplotypes.
- Statistical phasing methods use probabilistic models to evaluate possible haplotype combinations. Some approaches use hidden Markov models, while others use related probabilistic or optimization-based methods. The algorithms estimate which haplotype pairs are most consistent with the observed genotypes and population haplotype patterns. Their accuracy is influenced by sample size, population structure, allele frequencies, genetic relatedness, and the availability of suitable reference panels. Rare variants and populations that are poorly represented in reference datasets may be phased less accurately than common variants in well-represented populations.
- Reference-panel-based phasing uses previously characterized haplotypes from large population datasets. These reference panels provide information about common allele combinations and can improve phasing accuracy, particularly for common variants. Resources such as the 1000 Genomes Project and other population reference datasets have been used in statistical phasing and imputation workflows. The suitability of a reference panel depends on its population composition, variant coverage, genome assembly, and relevance to the study population. Differences in ancestry and population history can reduce accuracy when a reference panel does not adequately represent the individuals being analysed.
- Several tools are widely used for phasing. SHAPEIT is a statistical phasing program designed for large-scale genotype datasets and population-based analyses. Eagle is another widely used tool for statistical phasing, particularly in large cohorts. Beagle supports phasing and genotype imputation, using population haplotype information to infer missing or uncertain genotypes. These methods are particularly useful when working with genome-wide genotype data or large multisample variant datasets.
- Read-based and long-read phasing tools include WhatsHap, HapCUT2, and related methods. WhatsHap can use read information to phase variants and supports workflows involving short and long sequencing reads. HapCUT2 reconstructs haplotypes from sequencing reads and other linkage information. Long-read phasing methods may also incorporate structural variant evidence or assembly information, depending on the workflow. The choice of tool depends on whether the available evidence consists of short reads, long reads, family genotypes, population-level data, or a combination of sources.
- A simplified example of read-based phasing with WhatsHap is shown below:
# Phase variants using aligned reads and a VCF file
whatshap phase \
–reference=reference.fa \
–output=phased.vcf \
variants.vcf.gz \
sample.bam
- The command uses the reference genome, an input VCF containing heterozygous variants, and an aligned sequencing file. The output contains phase information for variants that can be linked using the available read evidence. In practice, the BAM file should be sorted and indexed, and the VCF should be prepared with compatible reference coordinates and genotype information. Exact options may vary with the WhatsHap version and the sequencing technology.
- A simplified statistical phasing workflow using SHAPEIT may look like this:
# Example structure for statistical phasing with SHAPEIT
shapeit4 \
–input variants.bcf \
–map genetic_map.txt \
–region chr1 \
–output phased_chr1.bcf
- The example illustrates the main inputs: genotype data, a genetic map, a genomic region, and an output file. Actual command-line options depend on the SHAPEIT version and the required reference-panel or population settings. Genetic maps provide information about recombination rates and help statistical methods model the relationship between physical distance and genetic distance.
- Haplotype reconstruction extends phasing by constructing the sequences of alleles carried on individual chromosomes. Once variants have been phased, the alleles assigned to each chromosome can be combined with the reference sequence to generate alternative haplotype sequences. This process can be useful for reconstructing genomic regions, identifying inherited haplotypes, analysing compound heterozygosity, and investigating the functional effects of multiple variants occurring together.
- For example, consider an individual with two heterozygous variants in a gene. If the variants are in trans, each chromosome carries one alternative allele, meaning that both gene copies may be affected by different variants. If the variants are in cis, both alternative alleles occur on one chromosome, while the other chromosome carries the reference alleles. This distinction can be clinically important in recessive disorders, where disease may result when pathogenic variants affect both copies of a gene. However, the clinical interpretation also depends on the pathogenicity of each variant, the disease mechanism, inheritance pattern, and relevant clinical evidence.
- Haplotype reconstruction can also help identify compound heterozygosity. Compound heterozygosity occurs when an individual carries two different variants in the same gene on opposite homologous chromosomes. Establishing that the variants are in trans supports this interpretation. If both variants are in cis, they occur on the same chromosome and the other copy may remain unaffected. Phasing is therefore an important part of analysing recessive disease-associated variants, although phase alone does not establish whether a variant is pathogenic.
- Haplotypes are also useful in population genetics and evolutionary studies. They can reveal patterns of genetic diversity, population structure, migration, recombination, and shared ancestry. Closely related individuals may share long haplotype segments inherited from common ancestors. The length and frequency of shared segments can provide information about relatedness and demographic history. Haplotype-based analyses are also used in studies of natural selection, founder effects, and population-specific genetic variation.
- In genome-wide association studies, haplotypes can help identify combinations of alleles associated with complex traits and diseases. Some genetic effects may depend on the combination of variants on the same chromosome rather than on individual variants considered independently. Haplotype association analysis can therefore provide complementary information to single-variant association testing. However, statistical associations do not automatically demonstrate causal effects, and analyses must account for population structure, linkage disequilibrium, relatedness, and multiple testing.
- Phasing is also important in pharmacogenomics. Multiple variants within a pharmacogenes can form distinct haplotypes, sometimes called star alleles. These haplotypes may be associated with differences in drug metabolism, transport, or response. For example, combinations of variants in genes such as CYP2D6, CYP2C19, and CYP2C9 can influence inferred gene function and metabolizer phenotype. Accurate haplotype assignment can therefore be important for pharmacogenomic interpretation. Complex structural variation, gene deletions, duplications, hybrid genes, and highly similar paralogous regions can make pharmacogene phasing particularly challenging.
- Haplotype reconstruction also supports genome assembly and structural variant analysis. Phased genome assemblies attempt to separate the two homologous chromosome sequences into distinct haplotypes rather than collapsing them into a single consensus sequence. This is particularly useful for diploid organisms with substantial heterozygosity. Long reads, linked reads, parental information, and sequencing coverage can all contribute to haplotype-aware assembly. Tools such as hifiasm and other haplotype-aware assemblers can produce phased or partially phased assemblies, depending on the input data and workflow.
- Several factors influence phasing accuracy. Sequencing depth determines how many reads support each variant linkage. Read length determines how many variants can be connected directly. Sequencing errors and mapping errors can create false allele combinations. Low-complexity and repetitive regions can cause reads to align to incorrect locations. Statistical methods may be affected by population structure, rare variants, and unsuitable reference panels. Trio-based approaches may be affected by incorrect family relationships or genotyping errors. These sources of uncertainty should be considered when interpreting phased genotypes and reconstructed haplotypes.
- Recombination is another important biological factor. During meiosis, homologous chromosomes exchange DNA segments, producing chromosomes with combinations of maternal and paternal genetic material. Recombination can divide long inherited haplotypes into shorter segments over generations. Statistical phasing methods use genetic maps and recombination models to account for these events. However, recombination hotspots and regions with unusual recombination patterns can affect haplotype reconstruction and the interpretation of shared segments.
- Phasing quality can be evaluated using several measures. Switch error rate measures the proportion of adjacent heterozygous sites for which the inferred phase relationship is incorrect. Haplotype block length indicates how much of the genome has been phased continuously. Phase completeness measures the proportion of informative heterozygous variants assigned to phase blocks. Accuracy can also be assessed by comparison with trio-based phase information, high-confidence reference haplotypes, or experimentally validated data. These measures provide complementary information, and no single metric fully describes the quality of a phased dataset.
- When evaluating phasing results, researchers should distinguish between phase accuracy and phase completeness. A method may correctly phase a small number of variants but leave much of the genome unphased. Another method may produce long phase blocks but contain occasional switch errors. Phase-set boundaries indicate where the method has not established a continuous relationship between allele assignments. They should not be interpreted as proof that biological recombination occurred at those exact positions.
- For reproducible haplotype analysis, researchers should preserve the original genotype data, reference genome assembly, sample and pedigree information, sequencing technology, read alignment files, genetic maps, reference panels, and software versions. The method used for phasing should be documented, including whether it relied on sequencing reads, parental genotypes, population statistics, or a combination of evidence. The output should be checked for phase-set consistency, missing genotypes, unexpected allele assignments, and possible errors in sample identity or reference assembly.
- Phasing and haplotype reconstruction are fundamental components of modern genomic analysis because they move beyond identifying which alleles are present to determining how those alleles are organized on individual chromosomes. Read-based methods, long-read sequencing, family-based approaches, and statistical algorithms provide complementary ways to infer haplotypes. Tools such as WhatsHap, HapCUT2, SHAPEIT, Eagle, and Beagle support different types of phasing workflows. Accurate phasing improves the interpretation of inherited variation, compound heterozygosity, population structure, pharmacogenomics, genome assembly, and disease-associated haplotypes. Understanding the evidence, assumptions, limitations, and quality measures behind these methods is essential for reliable genomic research.