Bayes' Theorem in Genetics — Use Bayes' theorem in pedigree/genetic screening contexts (intro) - Genetics

SECTION 1

Historical Context & Motivation

ℹ

How to Use This Lesson

This lesson is designed for a wide range of learners, so it is organized in two tiers. Grades 6–9 (Introductory Track): Focus on Sections 1–3 and the conceptual practice problem (Problem 1). These sections build the core ideas — prior, conditional, and posterior probability — using everyday analogies and visual diagrams, without requiring algebra beyond basic fractions. Grades 10–12 (Full Track): Work through all sections, including the mathematical framework (Section 4), the multi-step pedigree setup (Section 5), the full worked example (Section 6), and all practice problems. Prerequisite skills include comfort with fractions, basic algebra, and the idea of probability as a number between 0 and 1.

Imagine a family learns that a genetic condition, like cystic fibrosis, runs in their family tree. A family member wants to know: "What are my chances of being a carrier?" At first, you might guess based on simple Mendelian ratios. But what if that person has three healthy children — does that change the odds? This is exactly the kind of question that Bayes' theorem helps us answer. It is a mathematical tool that lets us update probabilities when new evidence appears.

The theorem is named after Reverend Thomas Bayes, an 18th-century English minister and mathematician. Bayes never published his famous idea during his lifetime — a friend sent it to the Royal Society after Bayes died. Over the centuries, scientists in many fields adopted the theorem. Geneticists found it especially powerful because family pedigrees constantly give us new clues that should shift our predictions.

1763 • Bayes' Essay Published Posthumously

Thomas Bayes' work on conditional probability is published by Richard Price, introducing the idea that prior beliefs can be updated with evidence.

1865 • Mendel's Laws of Inheritance

Gregor Mendel discovers the basic rules of how traits pass from parents to offspring, giving geneticists the prior probabilities they need.

1950s • Genetic Counseling Emerges

Doctors begin applying Bayesian reasoning to help families estimate the risk of inheriting genetic disorders based on family history and test results.

2000s–Today • Genomic Screening & Modern Diagnostics

Bayes' theorem becomes central to interpreting DNA tests, newborn screening panels, and carrier testing, where false positives and false negatives must be weighed carefully.

The core question Bayes' theorem answers in genetics is this: Given what we already know from a pedigree AND some new piece of evidence, what is the updated probability that a person has a certain genotype? Let's explore how this works step by step.

SECTION 2

Core Principles & Definitions

Before diving into the math, you need to understand a few key ideas. Bayes' theorem rests on the concept of conditional probability — the chance of something happening given that something else is already true. For example, the probability that someone is a carrier given that they have three unaffected children is a conditional probability.

Prior Probability

Your initial estimate of a genotype before considering new evidence. In genetics, this usually comes from Mendelian ratios or pedigree analysis. For example, a 2/3 chance of being a carrier based on parents' genotypes.

Conditional Probability (Likelihood)

The probability of observing your new evidence if a particular genotype were true. For instance, if a person IS a carrier, what is the chance all their children appear unaffected?

Posterior Probability

The updated probability of the genotype after combining the prior with the new evidence. This is the final answer Bayes' theorem gives you — a refined, more accurate estimate.

Pedigree (Family Tree)

A diagram showing how a trait passes through generations. Circles represent females, squares represent males. Filled-in shapes mean the person is affected by the condition. Pedigrees supply the prior probabilities for Bayesian analysis.

✦

KEY TAKEAWAY

Think of Bayes' theorem like a weather forecast that improves during the day. In the morning (prior), you hear a 60% chance of rain. But by noon you see blue skies (new evidence). You mentally update your estimate — maybe now it's only 20% (posterior). Bayes' theorem does this updating process with exact math instead of gut feeling. In genetics, the "morning forecast" is the Mendelian ratio, and the "blue skies" might be healthy children or a negative test result.

SECTION 3

How Bayes' Theorem Updates Probability — A Visual

The diagram below shows how Bayes' theorem works in a genetics scenario. A couple who are both carriers of an autosomal recessive condition (genotype Aa) has a child. Mendelian genetics predicts a 1/4 chance of being affected (aa), a 2/4 chance of being a carrier (Aa), and a 1/4 chance of being homozygous normal (AA). Among the unaffected children, 2/3 are carriers and 1/3 are homozygous normal — these are our prior probabilities. Then we add new evidence: the child has three unaffected children of their own (with a known carrier partner). The diagram shows how the posterior probability is calculated.

💡

Scaffolding Note

Grades 6–9 — Reading the diagram: Focus on the top row (Prior) and the bottom box (Posterior). The key insight is simply that the probability changed — from 67% down to 46% — because of the new information about the three healthy children. You do not need to follow every calculation step to understand the concept.

This flowchart traces a Bayesian update. The prior probability (top row) comes from Mendelian ratios. The conditional probability (middle row) asks how likely the new evidence is under each hypothesis. The joint probabilities are multiplied out, and the posterior is found by dividing each joint by their sum.

Notice how the evidence changed our estimate. Before knowing about the children, we said there was a 67% chance of being a carrier. After learning about three healthy children, that dropped to about 46%. The evidence didn't prove anything for certain — the person might still be a carrier — but it shifted the balance of probability. This is the power of Bayesian reasoning: every piece of evidence nudges the probability in one direction or the other.

SECTION 4

The Mathematical Framework

ℹ

Prerequisite Check

Grades 10–12 focus: Sections 4–6 introduce the algebraic formula and multi-step table method. You should be comfortable multiplying and dividing fractions, and understand what a fraction like 11/18 means as a probability before continuing. If you need a refresher on fraction arithmetic, review that before working through the formula below.

Now let's look at the formula itself. Don't worry — we will break it down piece by piece. In genetics, we usually apply Bayes' theorem to two competing hypotheses: a person either has genotype A or genotype B. We want to find the probability of genotype A given some observed evidence E.

BAYES' THEOREM (GENERAL FORM)

P(A | E) = P(A) × P(E | A) / [ P(A) × P(E | A) + P(B) × P(E | B) ]

P(A | E) = posterior probability of genotype A given evidence E. P(A) = prior probability of genotype A. P(E | A) = conditional probability of seeing the evidence if genotype A is true. P(B) = prior probability of the alternative genotype B. P(E | B) = conditional probability of seeing the evidence if genotype B is true.

In genetics, many counselors use a handy table method instead of plugging into the formula directly. You set up columns for each possible genotype and rows for the prior, the conditional, and the joint probability. Then the posterior is just the joint for your genotype divided by the total of all joints. This table approach makes even complex problems manageable.

POSTERIOR VIA TABLE METHOD

Posterior(A) = Joint(A) / [ Joint(A) + Joint(B) ]

Where Joint(A) = Prior(A) × Conditional(A) and Joint(B) = Prior(B) × Conditional(B). This is mathematically identical to the formula above but easier to organize.

CONDITIONAL FOR UNAFFECTED CHILDREN (AUTOSOMAL RECESSIVE)

P(n unaffected children | carrier × carrier) = (3/4)ⁿ

When two carriers of an autosomal recessive condition have children, each child has a 3/4 chance of being unaffected. For n independent children, multiply (3/4) by itself n times. If the person is NOT a carrier but their partner is, all children will be unaffected, so the conditional is 1.

ℹ

Why Does This Matter?

Without Bayes' theorem, genetic counselors would be stuck with the initial Mendelian ratios. They couldn't incorporate real-world observations like test results or the health of existing children. Bayesian reasoning lets them give families personalized, evidence-based risk estimates instead of one-size-fits-all numbers.

SECTION 5

Setting Up a Bayesian Pedigree Problem

Let's see how to use Bayes' theorem with a real pedigree. The diagram below shows a family with an autosomal recessive condition. Individual III-1 wants to know: what is the probability she is a carrier? Her grandparents (I-1 and I-2) are both carriers. Her father (II-1) is unaffected, and her uncle (II-3) is affected. She has married a known carrier and has had two unaffected children.

This pedigree shows an autosomal recessive condition across four generations. Individual III-1 is the person of interest. Her father (II-1) is unaffected but could be either AA or Aa. She has married a known carrier (III-2) and has two unaffected children. We will use Bayes' theorem to find her updated carrier probability.

To set up this problem, we need to work in two stages. Stage 1: figure out whether III-1's father (II-1) is AA or Aa. Since his brother II-3 is affected (aa), we know both grandparents are Aa. Among unaffected offspring of Aa × Aa, the prior probability of being Aa is 2/3 and of being AA is 1/3. Stage 2: determine III-1's genotype probabilities, then update them using her two unaffected children as evidence.

The combined prior for III-1 being a carrier (Step 4 in the table below) requires a two-branch calculation. We must consider both possible genotypes of her father and weight each by its probability: (1/3)(1/2) + (2/3)(2/3). Converting to a common denominator of 18 gives 3/18 + 8/18 = 11/18. This weighted-average approach is standard in multi-generation Bayesian pedigree problems and is the kind of calculation expected in upper high school genetics courses.

Step	Description	Values
1	Determine father's prior genotype probabilities (II-1 is unaffected from Aa × Aa)	P(AA) = 1/3, P(Aa) = 2/3
2	If father is AA: probability III-1 is a carrier (AA × Aa)	1/2
3	If father is Aa: probability III-1 is a carrier (Aa × Aa)	2/3 (among unaffected)
4	Combined prior P(III-1 is carrier) — weighted average across both father genotypes	(1/3)(1/2) + (2/3)(2/3) = 3/18 + 8/18 = 11/18
5	Apply Bayes' theorem using 2 unaffected children as evidence	See worked example in Section 6

Steps for setting up the Bayesian pedigree analysis of III-1

SECTION 6

Worked Example: Carrier Probability for III-1

Let's work through the full calculation for III-1 from the pedigree above. We want to find the probability that she is a carrier (Aa) given that she has two unaffected children with a known carrier (Aa) partner.

Finding the Posterior Probability That III-1 Is a Carrier

1
Step 1 — Determine the Prior Probability

III-1's father (II-1) comes from an Aa × Aa cross. Since he is unaffected, he is either AA (probability 1/3) or Aa (probability 2/3). Her mother (II-2) is a known carrier (Aa). If dad is AA, then the cross is AA × Aa, and the chance III-1 is a carrier is 1/2. If dad is Aa, then the cross is Aa × Aa, and among unaffected offspring, the chance of being a carrier is 2/3. So the combined prior for III-1 being a carrier is (1/3)(1/2) + (2/3)(2/3) = 1/6 + 4/9 = 3/18 + 8/18 = 11/18. The prior for NOT being a carrier is 1 − 11/18 = 7/18.

Prior: P(carrier) = 11/18 ≈ 0.611; P(non-carrier) = 7/18 ≈ 0.389
2
Step 2 — Determine the Conditional Probabilities

III-1 is married to a known carrier (III-2). They have 2 unaffected children. If III-1 is a carrier (Aa × Aa): Each child has a 3/4 chance of being unaffected. For 2 children: (3/4)² = 9/16. If III-1 is NOT a carrier (AA × Aa): No child can be affected (aa is impossible), so all children are unaffected with probability 1.

P(2 unaffected | carrier) = 9/16; P(2 unaffected | non-carrier) = 1
3
Step 3 — Calculate Joint Probabilities

Multiply the prior by the conditional for each hypothesis. Joint (carrier) = 11/18 × 9/16 = 99/288. Joint (non-carrier) = 7/18 × 1 = 7/18 = 112/288.

Joint(carrier) = 99/288; Joint(non-carrier) = 112/288
4
Step 4 — Compute the Posterior

The posterior probability of being a carrier is the carrier joint divided by the sum of both joints. Posterior = 99/288 ÷ (99/288 + 112/288) = 99/211 ≈ 0.469. The posterior probability of NOT being a carrier is 112/211 ≈ 0.531.

P(III-1 is a carrier | 2 unaffected children) = 99/211 ≈ 46.9%
5
Step 5 — Interpret the Result

Before considering her children, III-1 had about a 61.1% chance of being a carrier. After learning she has two unaffected children with a carrier partner, that probability dropped to about 46.9%. The evidence didn't eliminate the possibility, but it made it less likely. If she had even more unaffected children, the probability would keep decreasing.

Prior ≈ 61.1% → Posterior ≈ 46.9% (evidence shifted probability downward)

SECTION 7

Strengths & Limitations of Bayesian Genetics

Bayes' theorem is an incredibly useful tool, but like any tool, it has both strengths and limitations. Understanding these helps you know when to trust a Bayesian analysis and when to be cautious.

Strengths	Limitations
Incorporates new evidence — test results, number of healthy children, age of onset data — to refine risk estimates beyond simple Mendelian ratios.	Requires accurate prior probabilities. If the initial Mendelian ratios or population frequencies are wrong, the posterior will be wrong too.
Works with incomplete pedigrees where not everyone's genotype is known.	Assumes events are independent (e.g., each child's genotype is independent). This breaks down if there are complicating factors like genetic linkage.
Can combine multiple types of evidence: family history, biochemical tests, DNA tests, and clinical observations.	Not useful for traits controlled by many genes (polygenic traits), where simple Mendelian priors don't apply.
Gives a specific numerical probability rather than a vague yes/no answer, helping families make informed decisions.	Can be misunderstood if people confuse the posterior probability with certainty. A 10% carrier risk is not zero risk.

Strengths and Limitations of Bayes' Theorem in Genetics

✦

KEY TAKEAWAY

Think of Bayes' theorem as a GPS that recalculates your route whenever new information comes in. The GPS starts with a planned route (the prior), but if it detects a traffic jam (new evidence), it updates the estimated arrival time (the posterior). Similarly, every new observation — a healthy child, a test result — lets Bayes' theorem recalculate the genetic risk. The GPS is only as good as its maps, though, just like Bayes' theorem is only as good as its starting data.

SECTION 8

Connections to Advanced Genetic Analysis

The introductory Bayesian analysis you've learned in this lesson is a stepping stone to more sophisticated techniques used in modern genetics and medicine. As you study further, you'll encounter several extensions of this basic framework.

Introductory (This Lesson)	Advanced Extension
Two hypotheses: carrier vs. non-carrier for a single gene.	Multiple hypotheses: different possible genotypes at several loci simultaneously. Bayesian networks model complex inheritance patterns.
Evidence from offspring phenotypes (affected vs. unaffected).	Evidence from molecular DNA tests with known sensitivity and specificity (true positive and false positive rates). False positives must be carefully handled.
Autosomal recessive traits with simple dominance.	X-linked conditions, incomplete penetrance, variable expressivity, and age-dependent onset — each requiring modified conditional probabilities.
Manual calculation with a Bayesian table.	Computer algorithms using Markov Chain Monte Carlo (MCMC) methods to handle hundreds of family members and millions of genetic variants.

Introductory vs. Advanced Bayesian Genetics

One particularly important application is in genetic screening tests. When a newborn is screened for a metabolic disorder, the test might have a 99% detection rate but also a small false positive rate. Bayes' theorem helps doctors figure out: if a baby tests positive, what is the actual probability that the baby truly has the condition? This question is critical because rare conditions can have surprisingly high false positive rates relative to the number of true cases, a phenomenon known as the base rate fallacy. You'll explore this more as you advance in genetics and statistics.

💡

Looking Ahead

In future courses, you may encounter likelihood ratios and odds ratios — alternative ways to express Bayesian reasoning that are commonly used in clinical genetics. The core logic is exactly the same: start with what you know, incorporate new evidence, and arrive at a better estimate.

SECTION 9

Practice Problems

ℹ

Practice Problem Guide

Grades 6–9: Problem 1 (Conceptual) is accessible to all learners — it asks you to explain key ideas in your own words, with no calculation required. Grades 10–12: Attempt all five problems. Problems 2–5 require fraction arithmetic, multi-step Bayesian table calculations, and comfort with exponent notation such as (3/4)⁴. Work through each step systematically using the table method introduced in Section 4.

Question 1: CONCEPTUAL

In Bayesian genetics, what is the difference between a prior probability and a posterior probability? Explain in your own words why a genetic counselor would want to move from prior to posterior.

Show answer

A prior probability is the initial estimate of someone's genotype based only on Mendelian ratios and the pedigree structure — it's what you know before considering additional evidence. A posterior probability is the updated estimate after factoring in new information such as the phenotypes of offspring or the results of genetic tests. A genetic counselor moves from prior to posterior because the prior alone can be too general. For example, knowing that unaffected siblings of a carrier have a 2/3 chance of also being carriers is useful, but if that person has had four healthy children with a carrier spouse, the real risk is lower. The posterior gives a more personalized and accurate risk estimate, which helps families make better-informed decisions about future pregnancies or medical care.

Question 2: BASIC CALCULATION

Both parents are carriers (Aa) for an autosomal recessive condition. Their child is unaffected. Using the basic Mendelian prior, what is the probability that this unaffected child is a carrier (Aa)? Now suppose this child marries a known carrier and they have one unaffected child. Use Bayes' theorem to find the posterior probability that the original child is a carrier.

Show answer

From the Aa × Aa cross, the offspring ratio is 1/4 AA : 2/4 Aa : 1/4 aa. Since the child is unaffected (not aa), we exclude the 1/4 aa. Among unaffected children, the prior probability of being a carrier is 2/3, and of being AA is 1/3. Now for the Bayesian update: if the child is a carrier (Aa × Aa cross with their spouse), the probability of one unaffected child is 3/4. If the child is not a carrier (AA × Aa), the probability of one unaffected child is 1. Joint(carrier) = 2/3 × 3/4 = 6/12 = 1/2. Joint(non-carrier) = 1/3 × 1 = 1/3 = 4/12. Posterior(carrier) = (6/12) / (6/12 + 4/12) = 6/10 = 3/5 = 0.60 or 60%. The probability dropped from 66.7% to 60% after observing one unaffected child.

Question 3: INTERMEDIATE

A woman's brother has cystic fibrosis (autosomal recessive, genotype ff). Both of her parents must be carriers (Ff). She is unaffected. She marries a man who is a known carrier (Ff). They have four unaffected children. What is the posterior probability that the woman is a carrier?

Show answer

Since both parents are Ff and the woman is unaffected, her prior probability of being a carrier (Ff) is 2/3, and of being FF is 1/3. Now we apply Bayes' theorem with the evidence of 4 unaffected children. If she is a carrier (Ff × Ff cross): P(4 unaffected) = (3/4)⁴ = 81/256. If she is not a carrier (FF × Ff cross): P(4 unaffected) = 1. Joint(carrier) = (2/3)(81/256) = 162/768. Joint(non-carrier) = (1/3)(1) = 256/768. Total = 162/768 + 256/768 = 418/768. Posterior(carrier) = 162/418 = 81/209 ≈ 0.388 or about 38.8%. Having four unaffected children with a carrier partner substantially reduced her carrier probability from 66.7% down to about 38.8%.

Question 4: APPLIED

A genetic screening test for a rare autosomal recessive disease has a 95% detection rate (sensitivity) and a 2% false positive rate. The disease affects 1 in 10,000 people in the general population. If a randomly chosen person tests positive, what is the probability they actually have the disease? Use Bayes' theorem and explain why the result might be surprising.

Show answer

Let D = has disease and T+ = tests positive. Prior: P(D) = 1/10,000 = 0.0001. P(no disease) = 0.9999. Conditional: P(T+ | D) = 0.95 (sensitivity). P(T+ | no disease) = 0.02 (false positive rate). Joint(D) = 0.0001 × 0.95 = 0.000095. Joint(no D) = 0.9999 × 0.02 = 0.019998. Total = 0.000095 + 0.019998 = 0.020093. Posterior = P(D | T+) = 0.000095 / 0.020093 ≈ 0.00473 or about 0.47%. This means that even with a positive test result, there is less than a 1% chance the person actually has the disease! This is surprising because the test seems accurate (95% detection rate), but the disease is so rare that the false positives from the healthy population vastly outnumber the true positives from the tiny affected population. This is the base rate fallacy and illustrates why Bayes' theorem is essential for interpreting screening tests correctly.

Question 5: CRITICAL THINKING

Consider the pedigree from Section 5 (III-1 with a prior carrier probability of 11/18). Suppose instead of 2 unaffected children, III-1 has had 5 unaffected children with her carrier husband. Calculate the new posterior probability. Then discuss: could there ever be enough unaffected children to make the carrier probability exactly zero? Why or why not?

Show answer

Prior: P(carrier) = 11/18, P(non-carrier) = 7/18. With 5 unaffected children: If carrier (Aa × Aa): P(5 unaffected) = (3/4)⁵ = 243/1024. If non-carrier (AA × Aa): P(5 unaffected) = 1. Joint(carrier) = (11/18)(243/1024) = 2673/18432. Joint(non-carrier) = (7/18)(1) = 7/18 = 7168/18432. Total = 2673/18432 + 7168/18432 = 9841/18432. Posterior(carrier) = 2673/9841 ≈ 0.272 or about 27.2%. This is much lower than the original 61.1% prior — five healthy children provide strong evidence against carrier status. However, the posterior can never reach exactly zero. Mathematically, as long as the prior is not zero and the conditional probability (3/4)ⁿ is not zero (which it never is for any finite n), the joint for the carrier hypothesis remains positive. The posterior approaches zero as n approaches infinity but never actually reaches it. This reflects a real biological truth: no amount of unaffected children can completely rule out carrier status — only a direct DNA test could do that. Bayesian reasoning always allows for residual uncertainty, which is both a strength (it's honest about what we don't know) and a limitation (it never provides absolute certainty from indirect evidence alone).

SUMMARY

Lesson Summary

Summary

Bayes' theorem is a mathematical method for updating probabilities when new evidence appears. In genetics, it lets us go beyond simple Mendelian ratios by incorporating real-world observations. The process starts with a prior probability (the initial genetic estimate from a pedigree or population data), multiplies it by a conditional probability (how likely the evidence is under each hypothesis), and produces a posterior probability — the refined, evidence-based risk estimate. Each unaffected child, each test result, and each clinical observation shifts the probability toward a more accurate answer.

The practical applications are enormous. Genetic counselors use Bayes' theorem to advise families about carrier risks for conditions like cystic fibrosis and sickle cell disease. Doctors use it to interpret screening test results, where understanding false positive rates is critical. Remember that the theorem has limitations — it requires accurate priors, assumes independence of events, and works best for single-gene traits. But as a foundation for probabilistic reasoning in biology, Bayes' theorem is one of the most powerful tools in a geneticist's toolkit.