Dear Sandrine,
Thanks for pointing out this issue.
The BLAST databases we have on Galaxy are from last year, while those on
NCBI website are the latest (Jan 2012). As pointed out on NCBI website (
http://www.ncbi.nlm.nih.gov/Sitemap/sequenceIDs.html), it appears that each
time any change is made to a sequence/database, GI numbers change as well.
This is perhaps why you're observing discrepancies in GI numbers and
lengths between megablast outputs on Galaxy and NCBI. I'm currently in the
process of downloading the latest BLAST databases from NCBI, and I'll let
you know when they're available for use on Galaxy.
Thanks for your patience,
Guru
Galaxy team.
On Wed, Nov 9, 2011 at 8:03 AM, Sandrine Hughes <Sandrine.Hughes(a)ens-lyon.fr
wrote:
Dear all,
I’m not sure where I need to send my email so I apologize if I’m wrong.
I have a trouble with the Megablast program available in NGS Mapping and I
hope that you can help. Indeed, I think that there might be a problem with
the table given in output, and notably a shift between the GI numbers and
the parameters associated.
Here are the details:
I. First, what I have done :
I used the program to identify the species that I have in a mix of
sequences by using the following options:
Database nt 27-Jun-2011
Word size 16
Identity 90.0
Cutoff 0.001
Filter out low complexity regions Yes
I run the analyses twice and obtained exactly the same results (I used
the online version of Galaxy, not a local one).
II. Second, I analysed the data obtained for one of my sequence (1-202).
The following lines are the beginning of the table that I obtained after
the megablast and two lines with troubles:
1-202 312182292 484 99.33 150 1 0 1
150 1 150 2e-75 289.0
1-202 312182201 476 99.33 150 1 0 1
150 1 150 2e-75 289.0
1-202 308228725 928 99.33 150 1 0 1
150 19 168 2e-75 289.0
1-202 308228711 938 99.33 150 1 0 1
150 22 171 2e-75 289.0
1-202 308197083 459 99.33 150 1 0 1
150 10 159 2e-75 289.0
1-202 300392378 920 99.33 150 1 0 1
150 10 159 2e-75 289.0
1-202 300392376 918 99.33 150 1 0 1
150 9 158 2e-75 289.0
1-202 300392375 922 99.33 150 1 0 1
150 11 160 2e-75 289.0
1-202 300392374 931 99.33 150 1 0 1
150 21 170 2e-75 289.0
1-202 300392373 909 99.33 150 1 0 1
150 21 170 2e-75 289.0
1-202 300392371 1172 99.33 150 1 0 1
150 9 158 2e-75 289.0
...
1-202 179366399 151762 98.67 150 2 0 1
150 46880 47029 6e-73 281.0
1-202 58617849 511 98.67 150 2 0 1
150 21 170 6e-73 281.0
III. Third, what I’ve noticed:
My first trouble was that among all the species identified, two were
very different from the expected ones (2 last lines). So I decided to
search if that could be possible for that sequence and performed
independently a megablast on the NCBI with similar options. I was not able
to find these two species in the results.
So, I decided to check the hits identified in the table above and
identified a second trouble. In the table, the second column give the GI of
the database hit and the third column give the length of the database hit.
However, when I manually checked in NCBI the length of the GI, this one was
incorrect. Indeed, for the GI 312182292, the length should be 580 and not
484.
By checking different lines, I noticed that the length that is given
for a GI corresponds to the length of the GI-1. As you can see in the above
table, some GI are consecutive (300392376, 300392375,...). When checking
the length of 300392376 in NCBI, I should have 920. But when I checked
300392375, I found 918. And this was true for the following lines :
300392374 give normally 922 and 300392373 give 931... My conclusion at that
point was that there was a shift of –1 between the GI and the other
parameters of the line (indeed the parameters for the remaining columns are
in agreement with the length of the GI-1). However, that’s not always
true.... For some GI given in the table (for example, the two last lines),
if we check the parameters of the GI-1, the parameters are completely
different... So, I suppose that there is a trouble in the GI sorting during
the megablast but I’m not able to clearly define the problem.
IV. Fourth, confirmed with an other dataset
In order to be sure that the problem was not linked to my data or my
process, I asked a colleague to do a megablast on independent data. The
conclusions were similar to mine : a shift in the GI given in the table
and the parameters associated, that most of the time but not always,
correspond to GI-1.
Can you confirm that there is a problem with the output of the megablast
available in Galaxy ? If yes, do you think you can fix it ?
Many thanks for your help,
Best regards,
Sandrine
--
Graduate student, Bioinformatics and Genomics
Makova lab/Galaxy team
Penn State University
505 Wartik lab
University Park PA 16802
guru(a)psu.edu