What Penetrance Measures
A genotype does not always determine whether a phenotype is observed. Even people with the same genotype may differ in whether they express a trait, and the same person may or may not meet a phenotype definition depending on when and how it is assessed. Penetrance describes this relationship as a probability: among individuals with a specified genotype, how likely is the specified phenotype?
To make that question precise, define \(G\) as genotype and \(Y\) as an indicator for a particular phenotype: \(Y=1\) means the phenotype is present, and \(Y=0\) means it is absent. The phenotype must be defined before calculating penetrance. For example, “has the specified clinical feature by age 40” is more precise than simply “is affected.”
The conditioning matters. Penetrance starts with a genotype group and asks about phenotype within that group. It is not the proportion of all people with the phenotype who have that genotype, and it is not the frequency of the genotype in a population. Those quantities answer different questions.
Complete and Incomplete Penetrance
A genotype has complete penetrance for a specified phenotype when every individual with that genotype expresses that phenotype under the conditions being considered. In probability notation, its penetrance is \(p_g=1\). With incomplete penetrance, some individuals with the genotype do not express the phenotype, so \(p_g<1\). A genotype may have high, but still incomplete, penetrance.
These labels refer to the relationship between a specified genotype and a specified phenotype. They do not mean that a genotype is generally “penetrant” or “nonpenetrant” for every possible trait. The age at assessment and the definition of the phenotype matter: a phenotype that appears later in life may be absent in a younger individual who would express it at an older age. Accordingly, penetrance should be stated with enough context to identify what counts as expression and when it is assessed.
Complete penetrance is a claim about the underlying probability, not merely a description of a sample. If every observed individual with a genotype has the phenotype, the sample proportion is one. That observation does not prove that the population probability is exactly one: a finite sample may fail to include any of the individuals who would not express the phenotype.
Estimating Penetrance From Genotype-Specific Counts
Suppose a study records genotype and a binary phenotype for each individual. For a genotype \(g\), let \(n_g\) be the number of observed individuals with that genotype and \(x_g\) the number of them with the phenotype. The remaining \(n_g-x_g\) individuals have the phenotype absent. The empirical, genotype-specific estimate of penetrance is the affected count divided by the total count for that genotype:
This calculation requires \(n_g>0\); if no individuals with genotype \(g\) are observed, the data provide no sample proportion for that genotype. The estimate is between zero and one, inclusive. A value of zero means no sampled individual in that genotype group expressed the phenotype, while a value of one means all sampled individuals in that group did.
For this proportion to estimate the intended penetrance, the phenotype and genotype must be measured and classified consistently, and the sampled individuals within the genotype group must provide relevant information about the group of interest. If individuals are selected for study because they have the phenotype, for example, the observed fraction affected within a genotype group need not represent penetrance in the population. The data must include the relevant affected and unaffected individuals, not only cases.
Worked Example: Estimating Penetrance for Three Genotypes
In a hypothetical sample, a specified phenotype is recorded as present or absent, and genotype is classified as \(AA\), \(Aa\), or \(aa\). The phenotype definition and assessment age are the same for everyone. The counts are:
| Genotype | Phenotype present | Phenotype absent | Total |
|---|---|---|---|
| \(AA\) | 18 | 2 | 20 |
| \(Aa\) | 7 | 13 | 20 |
| \(aa\) | 0 | 20 | 20 |
For \(AA\), \(x_{AA}=18\) and \(n_{AA}=18+2=20\). Therefore,
The estimated penetrance of the specified phenotype for genotype \(AA\) is 0.900, or 90.0%, in the population represented by this hypothetical sample.
For \(Aa\), the calculation is
For \(aa\), it is
Thus, the observed proportions are 0.900, 0.350, and 0, respectively. These are estimates from the observed counts. In particular, the zero in the \(aa\) group does not establish that the underlying penetrance is exactly zero; it says that none of the 20 sampled individuals with genotype \(aa\) had the phenotype.
Why the Count Formula Estimates a Probability
The count formula has a simple justification. For the \(n_g\) observed individuals with genotype \(g\), let \(Y_i\) equal 1 if individual \(i\) has the phenotype and 0 otherwise. If, conditional on genotype \(g\), each individual has the same phenotype probability \(p_g\), then each \(Y_i\) has expected value \(p_g\). The sample proportion is the average of these indicators:
Taking its expected value and using the fact that the expected value of a sum is the sum of the expected values gives
So, under this common-probability model and appropriate sampling, the sample proportion is an unbiased estimate of penetrance: across repeated samples, its average equals the underlying probability. This result does not say that any one sample proportion must equal \(p_g\). The observed count can vary from sample to sample.
The equal-probability condition also clarifies why the phenotype definition and assessment conditions must be held fixed. If one part of a genotype group is assessed at a much younger age than another, or if “phenotype present” is classified differently between individuals, the observations may not represent one common probability \(p_g\).
Worked Example: A Sample Proportion of One
A hypothetical study assesses a clearly specified phenotype by age 30 in 15 individuals with genotype \(BB\). All 15 have the phenotype. The estimated penetrance is
The sample proportion is 1, or 100%. It is accurate to say that every sampled individual with genotype \(BB\) expressed the phenotype by age 30. It is not justified to conclude from these 15 observations alone that the underlying penetrance is certainly 1 or that the genotype has proven complete penetrance. The data are compatible with complete penetrance, but a larger relevant sample could include an individual with the genotype who does not express the phenotype.
The qualification “by age 30” is part of the interpretation. This estimate concerns the stated phenotype at that assessment age; it does not automatically describe expression at other ages or under a different phenotype definition.
Keep the Direction of Conditioning Straight
Penetrance asks for \(P(Y=1\mid G=g)\): the probability of the phenotype given a genotype. A frequent error is to reverse the condition and calculate \(P(G=g\mid Y=1)\), the proportion of phenotype-positive individuals who have that genotype. The two probabilities have different denominators and generally are not equal.
Worked Example: Penetrance Is Not the Genotype Mix Among Cases
In a hypothetical sample of 100 individuals, the genotype-specific counts are as follows:
| Genotype | Phenotype present | Phenotype absent | Total |
|---|---|---|---|
| \(CC\) | 12 | 28 | 40 |
| \(Cc\) | 18 | 42 | 60 |
For genotype \(CC\), the penetrance estimate is based on all 40 individuals with \(CC\):
For genotype \(Cc\), it is
By contrast, among the 30 phenotype-positive individuals, 12 have genotype \(CC\). The fraction \(12/30=0.400\) describes the proportion of cases with genotype \(CC\), not penetrance for \(CC\). Its denominator is all phenotype-positive individuals, whereas the penetrance denominator is all individuals with genotype \(CC\). Here the two genotype-specific penetrance estimates are both 0.300, even though 40.0% of the cases have genotype \(CC\).
Common Mistakes and Why Penetrance Matters
Penetrance provides a clear way to describe how often a specified phenotype occurs within a genotype group. It helps prevent a genotype from being treated as a guaranteed phenotype when expression is not universal. For interpretation to be useful, report the phenotype, genotype, assessment conditions, and counts alongside the estimated proportion.
- Using the wrong denominator. Dividing affected individuals with genotype \(g\) by all affected individuals gives the genotype composition among cases. Divide by all observed individuals with genotype \(g\) to estimate penetrance.
- Calling an observed 100% proof of complete penetrance. A sample proportion of one describes the sample. Complete penetrance is the claim that the underlying conditional probability is one.
- Leaving the phenotype vague. A penetrance estimate is tied to the phenotype definition and assessment conditions. State what counts as expression and, when relevant, the age of assessment.
- Confusing penetrance with genotype frequency. Genotype frequency describes how common a genotype is in a population. Penetrance describes the phenotype probability within a genotype group.
- Using case-only data as if they were a complete genotype group. Without the relevant phenotype-absent individuals, the denominator for a penetrance estimate is missing.
Check Your Understanding
Use the definitions and calculations above to answer these questions.
- In one sentence, define penetrance for a specified phenotype and genotype \(g\).
- A hypothetical sample contains 9 phenotype-positive and 21 phenotype-negative individuals with genotype \(Dd\). Calculate the estimated penetrance and state its interpretation.
- What does complete penetrance mean for a specified genotype and phenotype? Why does a sample proportion of one not prove it?
- Explain the difference between \(P(Y=1\mid G=g)\) and \(P(G=g\mid Y=1)\).
- A study includes only phenotype-positive individuals. What information needed to estimate penetrance within each genotype is missing?