[Bio-software] NCBI Blast Report - Expect(2)

jtmodel at gmail.com jtmodel at gmail.com
Wed Mar 22 19:03:14 EST 2006


Ah yes, a computer engineer trying to delve into the world of
bioinformatics.

Any advice or pointers to where to look for help would be much
appreciated (and if I get my stuff working, your stuff runs faster :).


After reading the Altschul papers, Ian Korf's great book, and more
tutorials and man pages than I care to mention, I'm seeing an output
from NCBI that I still can't quite figure out.

Specifically, mixed in with the E-values I understand,there appear to
be aberrant E-values labeled "Expect(2) n.nnn", where n is some number,
lower than what it should be for a given Raw Score, K,m,n,Lambda and H.

for example, here are two entries from the BLAST report...
(this was a single protein (LIM2) run against the drosophilia AA
sequence presented as one long string)
--------------------------------------------------------------------------
Score = 29.1 bits (55), Expect = 4.5
 Identities = 9/25 (36%), Positives = 14/25 (56%)

Query: 47      LGNKCYLQTDSIGEPPGQGPGRAWG 71
               L + C   T+ +G PPG  P ++ G
Sbjct: 3434216 LASSCGAATEQLGSPPGPSPSQSAG 3434240



 Score = 26.2 bits (49), Expect(2) = 0.068
 Identities = 7/16 (43%), Positives = 13/16 (81%)

Query: 14      VGTILLVVAMATDHWM 29
               V  ++ +VA++TDHW+
Sbjct: 6868559 VSLLIWIVAISTDHWI 6868574
----------------------------------------------------------------------------
Lambda     K      H
   0.331    0.141    0.483
-------------------------------------------------------------------------
The Raw scores appear to be "55" and "49", yet the Expect value for the
first HSP is MUCH higher than that for the second.

The calculation of the Expect value for the first sequence is good-old
E=(Kmn*exp(-Lam*S))/H

(And I don't really know how the factor of 1/H comes in, but it's
there).

For the second, I'm not really sure at all what Expect(2) means.  It
doesn't seem to relate to the E=(Kmn*exp(-Lam*S))/H equation at all.

Thanks for any input/insight,
--Josh Model
Boston University



More information about the Bio-soft mailing list