A3.5

Question 1 Single best answer Level 2 1 point

No solved relative of a target protein can be found. What do the columns of a family of its relatives give a predictor?

  1. They show which letters each position allows
  2. They average out errors in the target’s own sequence
  3. Their conserved positions mark which residues touch
  4. They point to a backbone the predictor can copy

Question 2 Single best answer Level 2 1 point

What did a protein language model such as ESMFold add over AlphaFold 2, and how does its accuracy compare?

  1. Folding from a deeper family search, and more accurately
  2. Folding from one sequence alone, but less accurately
  3. Folding from one sequence alone, and more accurately
  4. Folding from a deeper family search, but less accurately

Question 3 Single best answer Level 2 1 point

A lab has a deposited X-ray entry and a predicted model of the same protein. What does each one carry about how reliable its coordinates are?

  1. The entry, a per-residue confidence value; the model, no reliability value
  2. The entry, a resolution; the model, no reliability value
  3. The entry, a resolution; the model, a B-factor per residue
  4. The entry, a resolution; the model, a per-residue confidence value

Question 4 Single best answer Level 1 1 point

In an alignment of a protein family, which pattern marks coevolution?

  1. Two columns whose letters change together across the rows
  2. Two sequences whose letters change together across the columns
  3. Two neighboring positions in one sequence that change together
  4. Two columns that each hold one letter in almost every row

Question 5 Single best answer Level 3 1 point

Two positions in a target touch in the fold. The target’s family alignment holds a few different sequences. What will the alignment show for those two columns, and why?

  1. A clear correlation, because each change weighs more when rows are few
  2. Little or no correlation, because the pair changes in too few rows
  3. Little or no correlation, because touching positions stay conserved
  4. A clear correlation, because touching positions are under pressure to change together

Question 6 Single best answer Level 2 1 point

An arginine and an aspartate form a salt bridge (i.e., electrostatic attraction) in a protein’s fold. What pattern should their two columns show across the protein’s family alignment, and why?

  1. They change together, since swapping both charges keeps the bridge
  2. They change together, since both ends of the bridge share a gene
  3. They vary independently, since breaking the bridge at one end barely matters
  4. Both stay fixed, since changing either charge breaks the bridge

Question 7 Single best answer Level 2 1 point

In a measured structure, two positions lie far apart. Their alignment columns change together across the family. What best accounts for the correlation?

  1. Correlated columns mark touching residues, so the structure is wrong
  2. Each column is tied to a third, so the two correlate
  3. Each column allows few letters, so the two line up
  4. Both columns vary a lot, so the two vary together

Question 8 Single best answer Level 2 1 point

A predicted model gives residue 40 a confidence value. Which comparison would show whether that value was accurate?

  1. Residue 40’s position across repeated runs of the predictor
  2. The whole chain’s placement, aligned on a deposited structure
  3. Residue 40’s B-factor in a crystal entry of the protein
  4. Distances to its neighbors, against a deposited structure

Question 9 Single best answer Level 2 1 point

An AlphaFold file and an X-ray file of one enzyme both list a large value for residue 30 in the column headed B-factor. What does that value mean in each file?

  1. AlphaFold: likely well placed; X-ray: position smeared
  2. Both files: residue 30 is likely well placed
  3. Both files: residue 30’s position is smeared
  4. AlphaFold: position smeared; X-ray: likely well placed

Question 10 Single best answer Level 3 1 point

In a predicted model of one protein, the confidence values sit near 80 along both ends of the chain and near 25 along a stretch in the middle. What does the model support, going by these values?

  1. The fold of the middle, but not of either end
  2. The fold and placement of the ends, but not the middle
  3. The fold of each end, and the middle as a flexible linker
  4. The fold of each end, but not their placement

Question 11 Single best answer Level 3 1 point

Zippy is an invented predictor. At stated confidence 45, 65 and 90, its observed accuracy was 57, 65 and 78. How well calibrated is Zippy?

  1. Calibrated overall, as observed accuracy rises with stated confidence
  2. Overconfident at 45, calibrated at 65, underconfident at 90
  3. Calibrated overall, as the gaps at 45 and 90 offset each other
  4. Underconfident at 45, calibrated at 65, overconfident at 90

Question 12 Single best answer Level 2 1 point

A predicted model of G alone, an invented enzyme, has a low-confidence stretch. Few different sequences cover it, and in the cell a partner binds there. Which reasons stay possible, and what would test one?

  1. Few sequences, movement, a missing partner; predicting the complex tests the partner
  2. Movement, a missing partner; measuring G alone tests the partner
  3. Few sequences, a missing partner; predicting the complex tests the partner
  4. Few sequences, movement, a missing partner; measuring G alone tests the partner

Question 13 Single best answer Level 2 1 point

A co-folding model returns ranked samples of an invented protein and its ligand. Across the samples, one loop sits in several places. What does this spread tell you about the loop in the real protein?

  1. Not where it moves; the samples are tries at one answer
  2. Its main and minor states; the ranks show how common each is
  3. Where it moves and how often; the samples form an ensemble
  4. Where it moves, not how often; the samples carry no weights

Question 14 Single best answer Level 2 1 point

An invented enzyme works as a homodimer. A predictor given its sequence returns a model with high confidence over most of it. What does this model show of the working enzyme?

  1. One chain, plus a record of the dimer it forms
  2. Both chains, since the two share one sequence
  3. Both chains, apart from their low-confidence stretches
  4. One chain, without its partner or their interface

Question 15 Single best answer Level 4 1 point

A co-folding model returns a dimer of invented protein R. Both chains have confidence above 90 throughout, and the ipTM is low. Which of these analyses does the model serve: the fold of each chain, side-chain positions within each chain, and the residues that touch across the interface?

  1. Interface contacts only
  2. Chain folds, side chains and contacts
  3. Chain folds and side chains only
  4. Chain folds only
Back to top