From Variant Calls to Chromosome-Level Alleles
A genetic variant call describes what was observed at a position; it does not always say which variants lie together on the same copy of a gene. That distinction matters when representing genetic variation as a PBPK input: a subject’s two copies of a gene are naturally described as a pair of alleles, but an unphased list of calls may not determine that pair. This tutorial focuses on the assignment itself, not on what a particular allele does to enzyme activity.
At a gene locus, a person ordinarily has two homologous chromosome copies, one inherited from each biological parent. Each copy carries an allele of that gene. An allele can be described by the sequence across the positions of interest, or by a defined allele name when the observed variant combination matches an established definition. The name alone does not establish function: functional interpretation is a separate step.
The word “genotype” is used at more than one level. At a single position, a genotype states the two alleles observed there, such as A/G. Across several positions, a set of genotypes may report each position separately without revealing their chromosome-level arrangement. A diplotype adds that arrangement. In the examples below, square brackets list the alleles on one chromosome, and a slash separates the two homologs. The order of the homologs around the slash is arbitrary; swapping them does not create a different diplotype.
Homozygous, Heterozygous, and Compound Heterozygous
At one variant position, a person is homozygous if the alleles on the two homologs are the same at that position, and heterozygous if they differ. These terms describe a comparison between the two chromosome copies at a site. They do not by themselves describe the activity of the gene or the effects of a variant.
“Compound heterozygous” describes a more specific arrangement: two different variants in the same gene are present, with one variant on each homolog. In phase notation, the variants are in trans. If both variants are carried together on one homolog, they are in cis; the other homolog may carry neither of them. Finding two heterozygous calls is not enough to distinguish these arrangements. Phase must be established or the assignment must remain unresolved.
Worked Example: Reading a Phased Diplotype
Consider two positions in a hypothetical gene. At position 1, the reference allele is A and the alternate allele is G. At position 2, the reference allele is C and the alternate allele is T. A phased result is \([G,C]/[A,T]\). Identify the genotype at each position, state whether the subject is homozygous or heterozygous at each, and describe the phase.
The first chromosome carries G at position 1 and C at position 2. The second carries A at position 1 and T at position 2. Comparing the two homologs at position 1 gives G versus A, so the genotype there is A/G and is heterozygous. At position 2, the comparison is C versus T, so the genotype is C/T and is also heterozygous.
The phase is explicitly given: G is together with C, while A is together with T. Thus the diplotype is \([G,C]/[A,T]\). It would be incorrect to describe the subject as homozygous merely because each chromosome carries one allele at each position; homozygosity requires the two homologs to carry the same allele at the position being considered.
Why Unphased Calls Can Be Ambiguous
Suppose the only information is that a subject is A/G at position 1 and C/T at position 2. There are two distinct ways to distribute these alleles across the homologs. One is \([G,T]/[A,C]\), in which the alternate alleles are together in cis. The other is \([G,C]/[A,T]\), in which each homolog carries one alternate allele and the variants are in trans. Both arrangements produce exactly the same unphased genotype at each position.
The ambiguity can be counted for a simple case. Suppose there are \(k\) distinct biallelic positions, each heterozygous, and there is no phase information linking them. Fix one homolog as a reference for assigning the chromosome labels. At each position, either allele can be assigned to that homolog, giving \(2^k\) assignments. But swapping the labels of the two homologs describes the same diplotype, so the number of distinct diplotypes is \(2^k/2=2^{k-1}\), for \(k\geq 1\). This count assumes there are no additional constraints, such as a spanning read or known inheritance pattern. It is not a substitute for examining the evidence available in a particular case.
For two heterozygous positions, this gives \(2^{2-1}=2\) possible diplotypes, precisely the two arrangements above. If a position is homozygous, it does not create the same choice of which different allele goes on which homolog. Real data may also include missing calls, more than two possible alleles, or uncertainty in the calls themselves; these situations require care beyond this simple count.
Worked Example: Two Unphased Variants
In a hypothetical subject, sequencing reports A/G at position 1 and C/T at position 2 in the same gene. No reads span both positions, and no family data are available. List the compatible diplotypes and decide whether the variants are in cis or in trans.
At position 1, one homolog must carry G and the other A. At position 2, one homolog must carry T and the other C. Assigning G and T to the same homolog gives \([G,T]/[A,C]\), so the two alternate alleles are in cis. Assigning G and C to the same homolog gives \([G,C]/[A,T]\), so the alternate alleles are in trans.
Both diplotypes reproduce the reported calls: each has A/G at position 1 and C/T at position 2. Therefore, these calls alone do not establish whether the variants are in cis or trans. The conclusion is not “the variants are in trans” just because both sites are heterozygous; the phase is unresolved.
Evidence That Can Resolve Phase
Phase can be established using evidence that links variant observations to the same physical chromosome, or using inheritance information that identifies which alleles came from each parent. For example, a sequencing read or other molecule that spans two positions can show which alleles co-occur on that molecule. Such evidence must be evaluated against the assay’s coverage, error characteristics, and ability to distinguish the relevant chromosome copies. Family information can also constrain the assignments, provided the parental calls and inheritance assumptions are sufficiently clear.
A phase result should be reported only to the extent supported by the evidence. If an assay observes one variant on a molecule but does not span the other position, that observation does not link the two sites. Likewise, a list of separate site-level genotype calls does not become phased merely because the variants were detected in the same sample.
Worked Example: Using Spanning Reads to Support Phase
Return to the hypothetical two-position gene with reference alleles A and C and alternate alleles G and T. An assay produces 15 high-quality molecules that span both positions: 8 carry G and T, and 7 carry A and C. No molecules carry G and C or A and T. Assume the assay is designed so that each molecule reports the alleles on one chromosome copy, and that the calls are reliable for this illustration. What phase do these data support?
The spanning molecules directly link the positions. Eight molecules show G together with T, and seven show A together with C. The observed chromosome combinations are therefore \([G,T]\) and \([A,C]\), giving the supported diplotype \([G,T]/[A,C]\). The two alternate alleles are in cis.
The read counts also sum to the total number of molecules: \(8+7=15\). This checks the stated data, but the count alone does not guarantee that the assay is free of bias or error. The conclusion relies on the stated assumptions about spanning and reliable calls. If coverage were low, reads failed to span both positions, or allele dropout were plausible, the phase could be uncertain even if one arrangement appeared in the observed reads.
Worked Example: Identifying Compound Heterozygosity
A hypothetical gene has two distinct variants, v1 and v2. Variant calls show the reference and alternate allele at both sites. Independent phase evidence establishes \([\text{v1},\text{reference at v2}]/[\text{reference at v1},\text{v2}]\). Is this a compound-heterozygous arrangement?
The first homolog carries v1 and does not carry v2. The second homolog carries v2 and does not carry v1. Thus the two distinct variants are on opposite homologs: they are in trans. Under the definition used here, this is a compound-heterozygous arrangement in the gene.
If the phase evidence instead established \([\text{v1},\text{v2}]/[\text{reference at v1},\text{reference at v2}]\), both variants would be in cis. That would not be compound heterozygosity, even though the subject has a heterozygous call at each site. These descriptions concern arrangement only; they do not assert that either variant has a functional effect.
Diagnostics, Alternatives, and Limitations
A useful diagnostic is to reconstruct the site-level calls from every proposed diplotype. Each position should match the observed two alleles when the chromosome-level alleles are compared. If a proposed assignment fails that check, it is inconsistent with the calls. If two or more assignments pass, the calls do not uniquely determine phase.
Record the alleles observed at each position, keeping missing or uncertain calls distinct from reference alleles.
Group one allele from each position onto each homolog, using any direct or inheritance-based phase evidence.
Reconstruct each site-level genotype from the proposed diplotype. If several assignments remain consistent, report the ambiguity rather than choosing one without evidence.
Direct molecular phasing and inference from family data are alternative sources of phase information, not interchangeable guarantees. A read-based approach can link sites only when the relevant molecule spans them and the assay can distinguish the alleles reliably. Family-based inference depends on informative parental genotypes and correct assumptions about inheritance. Statistical or reference-panel inference may be used in other settings, but it adds model and population assumptions; it should not be confused with direct observation of the chromosome arrangement.
Several failure modes deserve attention. A variant may be missed because of inadequate coverage, or a call may be incorrect. Structural changes, copy-number differences, and variants outside the assayed region can complicate the assumption of exactly two fully observed homologs. An allele name may also depend on a specified definition or reference sequence; if the observed combination does not meet that definition, assigning the name can be misleading. These limitations do not change the meaning of phase, but they can prevent a confident diplotype assignment.
The central practical lesson is to keep three statements separate: which alleles were observed at each position, how those alleles are arranged across the two homologs, and what biological effect the resulting alleles may have. The first statement may be well supported while the second remains ambiguous. A PBPK model input should preserve that uncertainty rather than silently converting unphased calls into a single diplotype.
Common Mistake
A frequent error is to see two heterozygous variant calls and conclude that the subject has two variants on opposite chromosomes. That conclusion assumes a phase that the calls do not provide. With A/G at one site and C/T at another, both \([G,T]/[A,C]\) and \([G,C]/[A,T]\) fit the same unphased data. The correction is to identify the phase evidence, enumerate compatible diplotypes when needed, and state “unresolved” if the evidence does not distinguish them.
Check Your Understanding
Answer each question by using the allele arrangement, not by assuming that site-level calls reveal phase.
- A subject is A/G at one position. Is the subject homozygous or heterozygous at that position?
- At two positions, calls are A/G and C/T. Name both possible diplotypes using the notation in this tutorial.
- In \([G,T]/[A,C]\), are the alternate alleles at the two positions in cis or in trans?
- Two distinct variants in one gene are confirmed to be on opposite homologs. What arrangement term applies?
- Why does observing two heterozygous calls not, by itself, establish compound heterozygosity?
Solutions. (1) Heterozygous, because the two alleles differ. (2) \([G,T]/[A,C]\) and \([G,C]/[A,T]\); both reproduce the calls. (3) Cis, because G and T are on the same homolog. (4) Compound heterozygous, because the distinct variants are in trans. (5) The calls give alleles at each site but do not show which alleles share a chromosome; the variants could be in cis or in trans.