Rare Alleles And Approximate Carrier Frequencies

Why Approximate Carrier Frequency?

When a genetic variant is rare, a short calculation can often give a useful estimate of how many individuals carry it. Under Hardy–Weinberg equilibrium, the exact frequency of heterozygotes is \(2q(1-q)\), where \(q\) is the rare-allele frequency. If \(q\) is small, \(1-q\) is close to 1, so this frequency is often approximated by \(2q\). The approximation is convenient, but it is not exact: it overestimates the heterozygote frequency, and the size of that overestimate depends on \(q\).

The distinction matters when translating allele frequencies into expected carrier frequencies, or when deciding whether a simple approximation is accurate enough for a calculation. The appropriate question is not simply whether an allele is “rare,” but how large the approximation error is for the particular value of \(q\) and the accuracy required.

Exact and Approximate Frequencies

Consider a locus with two alleles. Write \(q\) for the frequency of the rare allele and \(p\) for the frequency of the other allele, so \(p+q=1\), or \(p=1-q\). We use “heterozygote carrier” here to mean an individual with one copy of the rare allele and one copy of the other allele. In a recessive-disease setting, that is often what “carrier” means; an individual with two copies of the rare allele is a different genotype.

Definition: Under Hardy–Weinberg equilibrium, the heterozygote carrier frequency for a rare allele of frequency \(q\) is \(2pq=2q(1-q)\). The rare-allele approximation replaces \(1-q\) with 1, giving \(2q\).

These are population genotype frequencies, equivalently the probabilities that a randomly selected individual has the specified genotype under the model. The calculation assumes the population at this locus is adequately described by Hardy–Weinberg genotype proportions for allele frequencies \(p\) and \(q\). In the standard idealized model, gametes unite at random with respect to the locus and have allele frequencies \(p\) and \(q\). This tutorial uses the resulting genotype proportions; it does not establish whether a particular population meets the model assumptions.

The term “rare” has no single cutoff that guarantees a particular accuracy. Instead, the formulas let us quantify the discrepancy directly. The approximation \(2q\) is always at least as large as the exact heterozygote frequency for \(0\leq q\leq 1\); for a nonzero rare-allele frequency, it is strictly larger.

Worked Example: A Very Rare Allele

Suppose a hypothetical population is in Hardy–Weinberg equilibrium and a rare allele has frequency \(q=0.005\), or 0.5%. Find the exact heterozygote carrier frequency, the \(2q\) approximation, and their difference.

The other allele has frequency \(p=1-q=1-0.005=0.995\). The exact heterozygote frequency is

$$ 2pq=2(0.995)(0.005)=0.00995. $$

As a percentage, this is \(0.995\%\). The approximation is

$$ 2q=2(0.005)=0.01000, $$

or \(1.000\%\). Thus the approximation exceeds the exact frequency by \(0.01000-0.00995=0.00005\), which is 0.005 percentage points. For this hypothetical allele frequency, the simple approximation is close to the exact Hardy–Weinberg value.

Deriving and Measuring the Error

The difference between the approximation and the exact frequency can be calculated without rounding either frequency:

2q-2q(1-q)=2q\bigl[1-(1-q)\bigr]=2q^2.

Therefore, the absolute error in frequency units is \(2q^2\). It is positive for \(q>0\), confirming that \(2q\) overestimates the exact heterozygote frequency. An error of 0.001 in frequency units is 0.1 percentage points, so when translating the error to a percentage it is important not to confuse percentage points with a relative percent error.

There are two useful ways to express the relative discrepancy, depending on which frequency is used as the denominator. Relative to the exact value, the overestimate is

\frac{2q-2q(1-q)}{2q(1-q)} =\frac{2q^2}{2q(1-q)} =\frac{q}{1-q}, \qquad 0<q<1.

Relative to the approximation \(2q\), the discrepancy is

\frac{2q-2q(1-q)}{2q} =\frac{2q^2}{2q} =q, \qquad q>0.

Both expressions describe the same difference; they use different denominators. When someone says the approximation “overstates the exact value by” a percentage, the usual denominator is the exact value, giving \(q/(1-q)\). If instead the error is expressed as a fraction of the approximation, it is \(q\). State the denominator when reporting relative error so that the percentage is unambiguous.

Key result: The exact heterozygote frequency is \(2q(1-q)\), and the approximation is \(2q\). The absolute overestimate is \(2q^2\); relative to the exact frequency, the overestimate is \(q/(1-q)\). Thus the relative error decreases as \(q\) decreases.

The relative-error formula also gives a direct tolerance rule. If the desired overestimate, measured relative to the exact frequency, is at most \(\varepsilon\), where \(\varepsilon>0\), then

\frac{q}{1-q}\leq\varepsilon \quad\Longleftrightarrow\quad q\leq\varepsilon(1-q) \quad\Longleftrightarrow\quad q(1+\varepsilon)\leq\varepsilon \quad\Longleftrightarrow\quad q\leq\frac{\varepsilon}{1+\varepsilon}.

This condition is precise: it lets the user choose a tolerated relative error first, then determine which allele frequencies satisfy it. For example, a 1% tolerance means \(\varepsilon=0.01\), not \(1\).

Worked Example: Checking a 1% Relative-Error Tolerance

A researcher is making a hypothetical calculation with \(q=0.008\) and wants the \(2q\) approximation to overstate the exact heterozygote frequency by no more than 1%, relative to the exact value. Check whether it meets this requirement.

For a 1% tolerance, \(\varepsilon=0.01\). The threshold for \(q\) is

q\leq\frac{0.01}{1+0.01} =\frac{0.01}{1.01} \approx 0.009901.

Since \(0.008<0.009901\), the specified frequency meets the tolerance. Direct calculation confirms this. Here \(p=1-0.008=0.992\), so the exact frequency is

$$ 2pq=2(0.992)(0.008)=0.015872, $$

while the approximation is \(2q=2(0.008)=0.01600\). The absolute error is \(0.01600-0.015872=0.000128\), and its relative size compared with the exact value is

\frac{0.000128}{0.015872} \approx 0.008065=0.8065\%.

The overestimate is below 1%, as required. This comparison is about the accuracy of the approximation under the assumed model; it is not a statement about uncertainty in an estimated allele frequency.

How Much Does the Approximation Matter?

The error grows with \(q\), both absolutely and relative to the exact frequency. The following hypothetical values show why the approximation should be checked rather than applied automatically.

Rare-allele frequency \(q\)	Exact \(2q(1-q)\)	Approximation \(2q\)	Absolute overestimate	Relative overestimate of exact value
0.005	0.00995 (0.995%)	0.01000 (1.000%)	0.00005 (0.005 percentage points)	0.503%
0.050	0.0950 (9.50%)	0.1000 (10.00%)	0.0050 (0.50 percentage points)	5.26%
0.200	0.320 (32.0%)	0.400 (40.0%)	0.080 (8.0 percentage points)	25.0%

For example, when \(q=0.05\), the relative overestimate of the exact frequency is \(q/(1-q)=0.05/0.95\approx0.0526\), or 5.26%. At \(q=0.20\), it is \(0.20/0.80=0.25\), or 25.0%. Calling an allele “rare” without stating the accuracy needed does not settle whether \(2q\) is suitable.

Worked Example: A Frequency Where the Approximation Is Poor

For a hypothetical population in Hardy–Weinberg equilibrium, suppose \(q=0.20\). Calculate and interpret both frequencies.

The other allele frequency is \(p=1-0.20=0.80\). The exact heterozygote carrier frequency is

$$ 2pq=2(0.80)(0.20)=0.32, $$

or 32.0%. The approximation gives

$$ 2q=2(0.20)=0.40, $$

or 40.0%. The approximation is higher by \(0.40-0.32=0.08\), which is 8.0 percentage points. Relative to the exact frequency, the overestimate is \(0.08/0.32=0.25\), or 25.0%. In this case, reporting \(2q\) as though it were the exact carrier frequency would be misleading for a calculation requiring close accuracy.

What “Carrier Frequency” Counts

The Hardy–Weinberg genotype frequencies are \(p^2\) for the homozygote with two copies of the other allele, \(2pq\) for the heterozygote, and \(q^2\) for the homozygote with two copies of the rare allele. In many recessive-disease discussions, “carrier” refers specifically to the heterozygote \(2pq\), while the rare homozygote is considered separately. That is the definition used for the \(2q\) approximation here.

If a question instead asks for the frequency of individuals with at least one copy of the rare allele, the rare homozygotes must also be included. That frequency is \(2pq+q^2=2q-q^2\), not \(2pq\). Its difference from \(2q\) is \(q^2\), rather than the \(2q^2\) error for the heterozygote-only frequency. The requested category must therefore be identified before selecting a formula.

Common Mistake

Wrong reasoning: “The rare-allele frequency is \(q\), so the carrier frequency is \(q\).” This treats the allele frequency as if it were the individual heterozygote frequency.
Correction: Under Hardy–Weinberg equilibrium, a heterozygote can receive the rare allele from either parental gamete: the genotype frequency is \(pq+qp=2pq=2q(1-q)\). When \(q\) is small, this is approximated by \(2q\), not \(q\).
Wrong reasoning: “The approximation \(2q\) is exact whenever an allele is called rare.”
Correction: Its absolute error is \(2q^2\), and its relative overestimate of the exact heterozygote frequency is \(q/(1-q)\). Compare that error with the required tolerance.
Wrong reasoning: “Carrier frequency always means the frequency with at least one rare allele.”
Correction: In common recessive-disease usage, a carrier often means a heterozygote. If the question includes rare homozygotes, calculate the frequency of that broader category explicitly.

Key takeaway: For a rare allele of frequency \(q\), the Hardy–Weinberg heterozygote carrier frequency is exactly \(2q(1-q)\). The convenient approximation \(2q\) overestimates it by \(2q^2\); use the relative error \(q/(1-q)\) to judge whether that overestimate is acceptable.

Check Your Understanding

Use the Hardy–Weinberg model for each calculation, and distinguish exact frequencies from approximations.

If \(q=0.01\), calculate the exact heterozygote frequency and the \(2q\) approximation. Give their absolute difference in frequency units.
Show that the relative overestimate of the exact heterozygote frequency is \(q/(1-q)\), and state what denominator defines this relative error.
For \(q=0.04\), calculate the relative overestimate of the exact frequency. Does \(2q\) meet a 5% tolerance measured relative to the exact value?
Explain why the frequency of individuals with at least one rare allele is not the same as the heterozygote carrier frequency.
For a 2% relative-error tolerance measured against the exact frequency, derive the largest \(q\) for which the \(2q\) approximation satisfies the tolerance.