Hidden Markov models for sequence: profiles, states and decoding · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.

One model for a whole protein family

Science · Genetics & Evolution · ages 22-23
Name ______________________   Date ____________

  1. A is aligned to consensus column 1 and T to column 3. The sequence has no residue at column 2. Which state skips that column?

    • The delete state
    • The match state
    • The insert state

  2. What does a profile HMM learn from a protein family's multiple alignment?

    • One representative sequence with the same gap costs at all positions
    • Position-specific emission and transition probabilities
    • Only the longest member, with no residue probability distributions

  3. A delete state emits extra residues between consensus columns.

    Circle one:   True   False

  4. A Viterbi path assigns insert state I2 to two residues. Their posteriors for I2 are 99% and 51%. What follows under this model?

    • The identical path labels give both residues equal posterior support
    • I2 has greater posterior support at the 99% residue
    • The 51% posterior makes the reported Viterbi path impossible

  5. Which method returns an allowed full state path with the highest probability for an observed sequence under a fixed HMM?

    • Independent marginal state selection
    • Ranking individual emission probabilities
    • Viterbi decoding

  6. A profile detects a remote match missed by a search with one family member. What can explain this?

    • It combines family-specific residue and gap patterns
    • It assigns equal emission scores to every residue at every family position
    • It treats all high-scoring sequences as proven to share an ancestor

  7. A variable profile column must assign equal emission probability to every residue.

    Circle one:   True   False

  8. A researcher reports only a Viterbi path and concludes that every residue's state assignment is certain. What information is missing?

    • A rule that makes all alternative paths impossible whenever a best path exists
    • A rule that declares all state assignments certain once a sequence is long
    • Posterior support for assignments across alternative paths
LightMySky · lightmysky.comW1-mt_nk14IaZTcw-s1

Answer key

For grown-ups. Fold this page away before handing over the rest.

One model for a whole protein family W1-mt_nk14IaZTcw-s1

  1. The delete state · D2 skips consensus column 2 without consuming or emitting a sequence residue.
  2. Position-specific emission and transition probabilities · The profile estimates residue emissions and transitions from aligned family members, rather than representing the family with one sequence.
  3. False · Insert states emit extra residues. Delete states skip consensus columns without emitting.
  4. I2 has greater posterior support at the 99% residue · The path labels agree, but the marginal support differs. The probabilities describe support under the sequence model.
  5. Viterbi decoding · Viterbi optimises a complete allowed path. Selecting states or emissions independently does not solve that full-path problem.
  6. It combines family-specific residue and gap patterns · Family-specific patterns can accumulate evidence across positions. This advantage is conditional and does not remove the need for significance checks.
  7. False · A variable column can allow many residues while assigning them different probabilities. Variability does not require a uniform distribution.
  8. Posterior support for assignments across alternative paths · A best path can have close alternatives that disagree at individual residues. State labels alone do not quantify that uncertainty.
Worksheet · LightMySky