A1.5

Question 1 Single best answer Level 1 1 point

In a BLAST result, what does a hit’s E-value report?

  1. The probability that one unrelated sequence would align this well in a single comparison
  2. The fraction of aligned positions at which the query and the hit carry the same residue
  3. The number of hits scoring this high that a database of this size would produce by chance
  4. The probability that the two aligned sequences are true homologs rather than a coincidence

Question 2 Single best answer Level 1 1 point

In a protein BLAST search, what is a word called when it differs from a query word but scores above the threshold T against it and so joins the seed list?

  1. A conserved motif
  2. A high-scoring segment pair
  3. A seed hit
  4. A neighbor word

Question 3 Single best answer Level 2 1 point

An exact database search would align the query against every database sequence using dynamic programming. Why is that approach too slow to run at the size of a modern sequence database?

  1. The number of possible alignments grows exponentially with length, so the matrix takes exponential time to fill
  2. Dynamic programming aligns both sequences end to end, so the whole matrix is filled even when the match is short
  3. Dynamic programming compares two sequences at a time, so the database has to be worked through one sequence after another
  4. Filling one matrix costs the query length times the sequence length, so the total work grows with the database

Question 4 Single best answer Level 2 1 point

BLAST cuts the query into short overlapping words and looks each one up in an index of the database. How does that lookup let BLAST avoid aligning the query against most of the database?

  1. Only the most conserved stretch of the query is used for the search, and the remaining positions are set aside
  2. The index stores the alignment score for each word pair, so scores are read from the table instead of aligned
  3. A sequence holding none of those words is skipped, and alignment grows outward only from the places the index flags
  4. Each database sequence is scored with a faster version of the alignment recurrence that visits every cell once

Question 5 Single best answer Level 3 1 point

A blastp search of a human query against the yeast proteome returns no significant hits, and you expect a homolog at about 30% identity. You rerun the search with a lower neighbor-word threshold T. What happens to the number of hits that get extended and to the runtime?

  1. More words pass the threshold, so more hits are extended and the search runs slower
  2. The seed list is unchanged, but each hit is extended further before stopping, so the search runs slower
  3. More words pass the threshold, so more hits are extended and the search runs faster
  4. Fewer words pass the threshold, so fewer hits are extended and the search runs faster

Question 6 Single best answer Level 3 1 point

Your query shares a 12-base regulatory motif with some distant relatives, and the sequence on either side of that motif is unrelated to theirs. To speed up a nucleotide search you raise the word size from 11 to 15. What happens to your chance of finding those relatives?

  1. The motif seeds as before, because word size only changes how many chance hits get extended
  2. The motif can no longer seed, because it is shorter than the exact match the search now requires
  3. The motif seeds less often, because longer words occur less frequently in the database
  4. The motif still seeds, but each extension stops sooner, so the alignment it produces is shorter

Question 7 Single best answer Level 4 1 point

A blastp search with a 310-residue query returns a hit at \(E = 2 \times 10^{-8}\) with 72% identity, over an alignment that covers only 78 residues of the query, and in both sequences that stretch is rich in proline and serine. What is the best interpretation of this hit?

  1. A composition artifact, because the score comes from shared residue bias rather than shared ancestry
  2. An unrelated match, because an alignment covering this little of the query cannot be trusted
  3. A real homolog, because an E-value this small cannot be produced by chance
  4. A shared functional domain, because high identity over a short conserved region is what a domain looks like

Question 8 Single best answer Level 4 1 point

You need a functional annotation for a 300-residue protein query. One hit shows \(E = 3 \times 10^{-9}\), 31% identity, and 95% query coverage. A second shows \(E = 1 \times 10^{-14}\), 96% identity, and 24% coverage over a region of ordinary composition. Which hit better supports annotating the whole query, and why?

  1. The first hit, because moderate identity across nearly the whole query fits a diverged relative
  2. The second hit, because its E-value is smaller and its identity higher, so it is the stronger match
  3. Neither hit, because identity in the low thirties is below the level at which homology can be claimed
  4. The second hit, because the short stretch of near-identity is the part that carries the function

Question 9 Single best answer Level 5 1 point

A search tool replaces BLAST’s contiguous 11-base word with a spaced seed: a match is required at 11 specified positions inside an 18-base window, in the pattern 111010010100110111, and the positions marked zero are ignored. Benchmarked against blastn at a word size of 11, the tool finds more weak homologs while examining about the same number of seed hits per megabase of database. Which explanation best accounts for both results together?

  1. The 18-base pattern is longer, so it is stricter, and cutting false seeds leaves more effort for real ones
  2. The gain comes from extending each hit further before the drop-off cutoff, rather than from the seed pattern
  3. The spaced rule is looser, so more seeds fire, and the added sensitivity is paid for with extra extension work
  4. The 11 required positions keep chance hits as rare, while the gaps let a region with scattered mismatches seed
Back to top