A question about blast algorithm

Question

[Denis]

I have some doubts regarding my understanding of blast algorithm that i'd like to dispel. It's well known that blast tries to find the same words in query and database and then uses it as starting point for the alignment extension. Let's imagine we have word AGTcCAT in query sequence and one word AGTgCAT in database. So there is a single mismatch between the words. My question, may the blast use this pair as starting point for the alignment extension? I think not, since blast requires only perfect matches of two words. Am i right? Thanks!

Answer 1

[Santosh Anand]

AFIAK, the seed region can't have mismatches. And the seed is created from the query string, which is matched against the database to find potential hits, which are then extended on both sides to get the final alignment. In your case, if the word size of seed region is >4, then you won't be able to find any hits. If seed <=3, there are matches (AGT = AGT; CAT=CAT etc), which can be extended according to mismatch penalty.

Note: For nucleotide, the defulat word-size is 11 and for AA, it is 3.

Answer 2

[Jean-Karim Heriche]

You're right if the word size is set correctly. See this explanation of the word size parameter.