Neural Sync Active

Learning Objectives

Registry Synced

Learning Objectives

378 words

2 min read

Reading compass

Now · 1. Why BLAST?

Learning Objectives

Understand BLAST heuristic algorithm
Interpret E-values, scores, and identity
Choose appropriate BLAST variants

1. Why BLAST?

Smith-Waterman is O(n*m) per query-database pair. Searching a database of 10^9 sequences would be impossible. BLAST uses heuristics for fast approximate search.

2. BLAST Algorithm Steps

Word generation: Break query into overlapping words (w=3 for protein, w=11 for DNA)
Seed identification: Find database matches to query words (allowing some mismatches)
Extension: Extend seeds in both directions to find High-scoring Segment Pairs (HSPs)
Evaluation: Report alignments with significant E-values

3. Interpreting BLAST Results

E-value: Number of alignments expected by chance with score >= S in database. E < 0.05 significant. Lower = more significant (less likely due to chance). Score: Bit score (normalized by scoring matrix statistics). Identity %: Percentage of aligned positions with identical residues. Query Coverage: Percentage of query length covered by alignment.

4. BLAST Variants

blastn: DNA query vs DNA database blastp: Protein vs protein blastx: Translated DNA vs protein database tblastn: Protein vs translated DNA database

Q1: Why can't we use Smith-Waterman for database searching?

Too slow - O(n*m) per pair. BLAST uses heuristics (word matching, seeding) to find significant alignments much faster while sacrificing some sensitivity. Q2: How to interpret E-value?

E-value = expected number of chance matches in database of given size. E=0.1 means 10% chance of seeing this score by chance. E < 0.05 = significant. Lower = more significant. Q3: How does BLAST find seeds?

Breaks query into overlapping k-mers (words). For protein: words of length 3. Searches database for identical or similar words (using BLOSUM threshold). These are seeds for extension. Q4: What is the difference between blastn and blastx?

blastn: DNA query vs DNA database (direct alignment). blastx: translate DNA query in 6 reading frames, then search protein database. Useful for finding protein-coding genes in genomic DNA. Q5: Why might a high-identity alignment have a poor E-value?

Short alignment length. E-value depends on both score AND length. A perfect match of 10 residues has high identity but short length, so may not be statistically significant (E-value > 0.05). Join Discord PreviousSequence AlignmentNextHidden Markov Models