[ssml] BLAST bug

[Mario Albrecht]

Dear all,

This is just to inform you about an interesting
(PSI-)BLAST bug, which leads to an incomplete sorting
of the search hits based on their E-values.
This problem may affect various bioinformatics methods
based on sorted (PSI-)BLAST output.

I reported this bug to the programming team in November,
but it seems that they could not fix it yet.
The bug occurs with different sequences and can already
occur after the initial search run, i.e. the first iteration.

For an example sequence to paste into the PSI-BLAST search of
http://www.ncbi.nlm.nih.gov/blast/, please see below.
Here, the bug occurs after the second iteration, just look
through the resulting list of search hits.
This list may also be obtained for a short while via the
following request ID: 1076612181-1356-65367656245.BLASTQ3

>sp|P25644
MSFFGLENSGNARDGPLDFEESYKGYGEHELEENDYLNDETFGDNVQVGTDFDFGNPHSS
GSSGNAIGGNGVGATARSYVAATAEGISGPRTDGTAAAGPLDLKPMESLWSTAPPPAMAP
SPQSTMAPAPAPQQMAPLQPILSMQDLERQQRQMQQQFMNFHAMGHPQGLPQGPPQQQFP
MQPASGQPGPSQFAPPPPPPGVNVNMNQMPMGPVQVPVQASPSPIGMSNTPSPGPVVGAT
KMPLQSGRRSKRDLSPEEQRRLQIRHAKVEKILKYSGLMTPRDKDFITRYQLSQIVTEDP
YNEDFYFQVYKIIQRGGITSESNKGLIARAYLEHSGHRLGGRYKRTDIALQRMQSQVEKA
VTVAKERPSKLKDQQAAAGNSSQDNKQANTVLGKISSTLNSKNPRRQLQIPRQQPSSDPD
ALKDVTDSLTNVDLASSGSSSTGSSAAAVASKQRRRSSYAFNNGNGATNLNKSGGKKFIL
ELIETVYEEILDLEANLRNGQQTDSTAMWEALHIDDSSYDVNPFISMLSFDKGIKIMPRI
FNFLDKQQKLKILQKIFNELSHLQIIILSSYKTTPKPTLTQLKKVDLFQMIILKIIVSFL
SNNSNFIEIMGLLLQLIRNNNVSFLTTSKIGLNLITILISRAALIKQDSSRSNILSSPEI
STWNEIYDKLFTSLESKIQLIFPPREYNVHIMRLQNDKFMDEAYFGQFLASLALSGKLNH
QRIIIDEVRDEIFATINEAETLQKKEKELSVLPQRSQELDTELKSIIYNKEKLYQDLNLF
LNVMGLVYRDGEISELK


Kind regards,

  Mario Albrecht