What is allele dosage?

Genotype Data

Ploidy & Dosage

Understand what allele dosage represents in diploids and polyploids, how it differs from read fraction, and why dosage uncertainty matters.

Author

Alexander M. Sandercock

Published

July 13, 2026

Keywords

allele dosage, ploidy, polyploid, genotype, read depth, dosage calling

Allele dosage is the number of copies of a specified allele carried by an individual at a locus. At a biallelic locus in an organism with ploidy \(P\), dosage can range from 0 to \(P\).

Start with a diploid example

Suppose a SNP has two alleles, A and B, and we define dosage as the number of copies of B.

Genotype	B-allele dosage
`AA`	0
`AB`	1
`BB`	2

In a diploid, the three genotype states correspond directly to dosages 0, 1, and 2.

What changes in a tetraploid?

A locus modeled at tetraploid ploidy has four chromosome copies contributing to the genotype. For the same A/B SNP, five dosage states are possible:

0AAAA

1AAAB

2AABB

3ABBB

4BBBB

The intermediate heterozygous states are biologically distinct. Collapsing all of them into a single AB class discards information that can matter for linkage mapping, association analysis, genomic prediction, and other quantitative analyses.

Dosage is allele-specific

A dosage value has meaning only after the counted allele is defined. If B dosage is 1 in a tetraploid, then A dosage is 3. When exchanging data, document whether dosage counts the reference allele, alternate allele, or another explicitly named allele.

Dosage is not the same as read fraction

For sequencing data, a sample with dosage 2 in a tetraploid has an idealized expected B-allele fraction of \(2/4 = 0.5\). But an observed read fraction of 0.5 is not automatically proof of dosage 2.

Observed allele counts are affected by:

finite read sampling, especially at low depth;
allele-specific amplification or capture bias;
sequencing error;
overdispersion beyond simple binomial sampling;
outlying observations;
mapping of reads from paralogous or homoeologous regions.

This is why dosage callers use the read evidence, and often population-level information, to estimate genotype probabilities rather than simply rounding the observed allele fraction.

Idealized expectation

Tetraploid dosage 1 suggests an alternate-allele fraction near \(1/4 = 0.25\) when both alleles are sampled equally.

Real observation

A dosage-1 individual might produce 3 alternate reads out of 20, 7 out of 20, or another count. The caller must decide how compatible that evidence is with each dosage state.

Dosage, ploidy, and copy number are different concepts

Term	Question it answers	Example
Ploidy	How many chromosome-set copies are being modeled?	Tetraploid: \(P=4\)
Allele dosage	How many copies of a specified allele occur at this locus?	`B` dosage = 3
Copy number	How many copies of a genomic region or sequence are present?	A segment has 5 copies

They can interact, but they are not interchangeable. For example, an aneuploid or copy-number-variable region can violate the assumption that every locus follows the organism-wide ploidy.

Why dosage uncertainty matters

A hard call such as 2 hides the fact that some calls are much more certain than others. Two samples can both be assigned dosage 2 while having very different read depth and posterior support.

For sequencing-based genotyping, check whether the workflow retains or exposes quantities such as:

posterior genotype probability;
genotype quality;
read depth;
allele bias estimates;
model-fit diagnostics.

When uncertainty is substantial, treating every hard dosage call as equally reliable can propagate error downstream.

Practical takeaway

For a biallelic locus at ploidy \(P\), code dosage consistently as an integer from 0 through \(P\), state which allele is being counted, and keep genotype uncertainty available whenever the calling method provides it.

Continue learning

Process MADC genotype data Estimate ploidy and copy-number variation in alfalfa