Films · Genetics and Molecular Biology

Evidence Reweights the Possibilities

Watch Evidence Reweights the Possibilities · 3:47
Read the transcript of “Evidence Reweights the Possibilities”

Film transcript: Evidence Reweights the Possibilities

The first film tested inheritance models against family observations. How can learning that someone is unaffected raise their carrier probability? The surprise comes from comparing only genotypes still compatible with the observation.

Begin with every genotype the inheritance model allows. Give each possibility its starting weight before using the person's phenotype. These weights come from transmission rules and any population assumptions. Write them down before looking at the answer you hope to get. Now condition on what you observed. Any genotype incompatible with that observation loses its weight. Compatible possibilities remain. This is a filter, not a change to the person's cells. Keep the survivors in their original proportions. Suppose a new observation excludes one visible possibility. Which tiles should remain in the probability bar, and what must happen next? Watch the total bar, not just the disappearing tile.

Remove the incompatible tile. Then stretch the survivors to fill the whole bar again. That second step is renormalizing. The surviving possibilities now divide all the probability between them. A carrier tile can therefore occupy more of the remaining bar. This can happen even when the person is unaffected. The comparison set became smaller. Nothing happened to the person's DNA.

Evidence changes genotype probabilities, not genotypes. Only the description of your uncertainty changed. Sometimes evidence does not eliminate a hypothesis. It is merely more likely under one hypothesis than another. Bayes' rule handles that case. The observation changes weights without making either hypothesis impossible. Likelihood measures that predictive fit. Each hypothesis starts with a prior weight. Multiply that weight by how well the hypothesis predicts the observation. A better prediction earns a larger updated weight. A weak prediction earns less. Finally, renormalize those weighted hypotheses. Their new shares form the posterior: the probabilities after accounting for the evidence. Together, those shares still fill the complete probability bar.

More observations repeat the same cycle. Start with current weights, ask what each hypothesis predicts, apply the evidence, then renormalize. Each new fact becomes the next filter. The method remains the same. The answer still depends on the model. Penetrance, new mutation, family relationships, and population assumptions all shape those starting weights and likelihoods. If an assumption changes, recompute instead of reusing the old answer. With that picture, the text can calculate exact carrier and recurrence probabilities. Keep the filter, the weights, and the assumptions visible. The worked examples begin with different family evidence.

Your brother has cystic fibrosis and you do not. Is your chance of being a carrier 1/2 or 2/3?

Continue in the lesson: Reading Pedigrees: Human Inheritance →

3:47 · Free, with captions and a transcript. The lesson has the full text, the questions and the sources.

More films from this lesson

How this film was made. Narration: an AI narrator voice (Kokoro, af_heart) reading narration written from this lesson's text. Animation: AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures. Captions come from the narration script. See something wrong? Report it.