Mendelian Segregation Ratios and Sampling Distributions

From Mendelian Transmission to Family Counts

A Mendelian segregation ratio describes probabilities for the alleles transmitted in a single meiosis. A family, however, contains a finite number of offspring, so its observed genotype counts need not match those probabilities exactly. A cross with an expected 1:1 ratio can produce more offspring of one genotype than the other simply through sampling variation. The sampling distribution quantifies how likely different counts are under the specified parental genotypes and transmission rules.

This tutorial considers autosomal genotypes at one locus, with two alleles \(A\) and \(a\), and focuses on crosses with at least one heterozygous parent \(Aa\). The goal is to derive the probabilities for one offspring, then use them to calculate probabilities for counts across \(n\) offspring. These are genotype-count calculations: a genotype does not by itself determine an affected or unaffected phenotype.

Offspring Genotype Probabilities

A heterozygous parent \(Aa\) transmits \(A\) with probability \(1/2\) and \(a\) with probability \(1/2\). The other parent contributes an allele according to that parent's genotype. Combining the possible transmitted alleles gives the offspring probabilities. For example, if the other parent is \(aa\), it always contributes \(a\): the offspring are \(Aa\) with probability \(1/2\) and \(aa\) with probability \(1/2\).

If both parents are heterozygous, each transmits \(A\) or \(a\) with probability \(1/2\). The two parental transmissions combine to produce \(AA\), \(Aa\), and \(aa\) with probabilities \(1/4\), \(1/2\), and \(1/4\), respectively. These probabilities form a 1:2:1 genotype ratio. If the other parent is \(AA\), the offspring are \(AA\) or \(Aa\), each with probability \(1/2\).

A ratio is not a promise about a particular family. For instance, the 1:1 ratio in an \(Aa \times aa\) cross gives equal probabilities for the two genotypes in each offspring; it does not require exactly half of every family to have each genotype. The count distribution describes that variation.

Assumptions for Count Distributions

The calculations below use a standard Mendelian sampling model. Parental genotypes are known and treated as fixed; each meiosis transmits either allele from a heterozygous parent with probability \(1/2\); and transmissions to different offspring are independent conditional on those parental genotypes. The offspring are drawn from the same cross, so their genotype probabilities are the same. The model also assumes no mutation at the locus, segregation distortion, genotype-dependent survival before observation, or genotyping error.

These assumptions concern genotype transmission and observation. If the data are phenotypes rather than genotypes, penetrance and phenocopy mechanisms are needed to connect genotype to phenotype. Those mechanisms are not part of the count distributions derived here. Likewise, if offspring have not been observed independently or the sampling process selects families based on their outcomes, the simple distributions may not describe the observed counts without additional modeling.

Binomial Counts for Two Categories

When the outcome of interest has two categories, such as \(Aa\) versus \(aa\) in an \(Aa \times aa\) cross, the count in one category has a binomial distribution. Let \(X\) be the number of offspring with a specified genotype among \(n\) offspring, and let that genotype have probability \(p\) for each offspring. Then \(X \sim \mathrm{Binomial}(n,p)\), with probability of exactly \(k\) such offspring given by:

The term \(p^k(1-p)^{n-k}\) is the probability of one particular ordering with \(k\) offspring in the specified category and \(n-k\) in the other category. There are \(\binom{n}{k}\) possible placements of the \(k\) offspring, which gives the formula. The expected count is \(np\), and the variance is \(np(1-p)\). Thus, even if the expected count is an integer, other counts generally have positive probability.

Multinomial Counts for Three Genotypes

When the outcomes are \(AA\), \(Aa\), and \(aa\), the counts must be considered jointly because they add to \(n\). Let the respective counts be \(n_{AA}\), \(n_{Aa}\), and \(n_{aa}\), with \(n_{AA}+n_{Aa}+n_{aa}=n\), and let the corresponding offspring probabilities be \(p_{AA}\), \(p_{Aa}\), and \(p_{aa}\). Under independent offspring outcomes, the probability of this exact count combination is:

The factorial coefficient counts the distinct orderings of the three genotype categories. Each particular ordering has probability \(p_{AA}^{n_{AA}}p_{Aa}^{n_{Aa}}p_{aa}^{n_{aa}}\), so multiplying by the number of orderings gives the probability of the count combination. This is the multinomial distribution. It requires nonnegative integer counts that sum to \(n\), and genotype probabilities that sum to one.

Counting a Group of Genotypes

Sometimes the question concerns a group of genotypes rather than each genotype separately. For example, in an \(Aa \times Aa\) cross, an offspring is a carrier of at least one \(A\) allele if its genotype is \(AA\) or \(Aa\). The probability of that grouped outcome is \(1/4+1/2=3/4\); the probability of \(aa\) is \(1/4\). Therefore, the carrier count among \(n\) offspring has a binomial distribution with \(p=3/4\). This grouping does not change the underlying genotype probabilities; it combines categories for a question that only distinguishes carrier from non-carrier.

Why the Derivation Matters

A short derivation shows why the same assumptions lead to both binomial and multinomial sampling distributions. For \(n\) offspring, conditional independence means the probability of any fixed ordered genotype sequence is the product of the individual genotype probabilities. If a sequence contains \(n_j\) outcomes in category \(j\), its probability is \(\prod_j p_j^{n_j}\). The number of distinct sequences with those counts is \(n!/\prod_j n_j!\). Multiplication gives the multinomial probability. For two categories, the coefficient and probability product reduce to the binomial formula.

If categories are combined, their probabilities add. For example, in an \(Aa \times Aa\) cross, the probability of \(AA\) or \(Aa\) is \(3/4\), leaving \(1/4\) for \(aa\). Each offspring then has a two-category outcome for the carrier-count question, which yields the binomial distribution. This reduction is valid because the question asks only for the total number in the combined group.

Common Pitfalls and Interpretation

The main distinction is between a theoretical ratio and a realized count. A 1:1 segregation ratio says that each offspring has probability \(1/2\) of each of two genotypes under the model. It does not say that every family will split evenly. Likewise, a 1:2:1 ratio in an \(Aa \times Aa\) cross is a set of single-offspring probabilities, not a required pattern in every small family.

• Wrong reasoning: “The cross has a 1:1 ratio, so eight offspring must include four of each genotype.” Correction: The expected counts are four and four, but the binomial distribution assigns probability to other counts as well.
• Wrong reasoning: “The observed count has probability 0.11, so the model has only an 11% chance of being correct.” Correction: That probability is calculated assuming the model; it is not a probability that the model is true.
• Wrong reasoning: “The multinomial probability for the exact genotype counts can be calculated without the ordering coefficient.” Correction: A family count does not specify which child had which genotype, so all distinct orderings consistent with the counts must be included.
• Wrong reasoning: “A carrier count can always be interpreted as an affected count.” Correction: Genotype and phenotype are distinct; a phenotype probability requires an appropriate penetrance and phenocopy model.

These distributions provide the probability calculations needed to describe variation in family segregation counts under fixed Mendelian transmission. Their interpretation depends on the cross, the event being counted, and the assumptions about transmission and observation. A numerical result is meaningful only when those elements are stated clearly.