Mysterious Lost Version Of Blast (2.2.20)

Question

[Science_Robot]

Some people in my lab use megablast from BLAST 2.2.20 claiming that it is magically faster than the latest.

So I went to search for it to see what the difference is:

...Deep in the dungeons of NCBI

> ftp> ls
229 Entering Extended Passive Mode (|||50259|)
150 Opening ASCII mode data connection for file list
dr-xr-xr-x   2 ftp      anonymous     4096 Nov 20  2008 2.2.18
dr-xr-xr-x   2 ftp      anonymous     4096 Jul 22  2009 2.2.19
dr-xr-xr-x   2 ftp      anonymous     4096 Jul 27  2009 2.2.21
dr-xr-xr-x   2 ftp      anonymous     4096 Oct 14  2009 2.2.22
dr-xr-xr-x   2 ftp      anonymous     4096 Mar 22  2010 2.2.23
dr-xr-xr-x   2 ftp      anonymous     4096 Nov 12 16:17 2.2.24
lr--r--r--   1 ftp      anonymous        6 Aug 23  2010 LATEST -> 2.2.24

What?! Where is 2.2.20?

Did the magi of NCBI retract it from their FTP for simply being too fast for the database queries of man?

Do any NCBI historians know what happened?

Answer 1

[Michael Schubert]

You were most likely looking in the ftp.ncbi.nih.gov/blast/executables/blast+ directory. According to the release notes, 2.2.21 was the first to ship the BLAST+ command-line tools (the older versions of BLAST+ were most likely not fully ported to C++ at the date of the release).

2.2.20 can be found the in the C-tree of the releases. Since you mentioned megablast, the version your people are using definitely belongs there.

Note that there was a discussion on Biostar whether BLAST or BLAST+ was faster that came to the opposite conclusion.

[Michael Schubert]

Well, see if your people are actually using 2.2.20 or 2.2.20+. If it's the first one, NCBI will not have released 20+ for some reason. If it's the latter and NCBI did take down the release again, they probably had a good reason to and I suggest you don't use it.

[Science_Robot]

But versions 2.2.19 and 2.2.18 are still there.

Also, I'm talking about the megablast executable for blastn. I think the difference might just be a result of default word-length.

Also, 2.2.20 uses all cores regardless of whether or not the subject database is split up.

[Michael Schubert]

Well, see if your people are actually using 2.2.20 or 2.2.20+. If it's the first one, NCBI will not have released 20+ for some reason. If it's the latter and NCBI did take down the release again, I suggest they had a good reason to and you don't use it.

[Science_Robot]

Are the C and C++ versions being developed in parallel? I noticed that there is a version 2.2.25 for the C version.