Is BLAST using Needleman-Wunsch or Smith-Waterman when aliging in between HSSPs?

Question

[Arjen Ten Have]

I am going thru my lectures for next month postgrad course and ran into a doubt. To me it would make sense to apply Needleman-Wunsch (Global) when aligning the subsequences in between the HSSPs identified in the earlier steps. But my class (which I made a couple of years ago) actually states it uses a "Smith-Waterman-like algorithm". Cannot find this in NCBIs handbook, Anybody any idea?

Answer 1

[Alex Reynolds]

A thorough walkthrough of the statistical method for an early version of BLAST can be found on pg. 345 of Ewens and Grant's Statistical Methods in Bioinformatics. Might be worth investigating.

Answer 2

[mfilipav]

BLAST uses a local alignment algorithm, namely, Smith-Waterman.

[Ram]

That seems a little broad/inaccurate. Would you happen to have a link to any documentation that suggests BLAST uses an algorithm as simple as SW?

[Arjen Ten Have]

Are you sure? I understood bLast is local since it is based on perfect, hence local matches that are extended to HSSPs that then become connected by a NW. It doesn't make sense to use SW since alignment between the HSSPs is from 0,0 to n,m where 0,0 corresponds with the first position following HSSPn and n,m is the position prior to HSSPn+1.