Crux error when searching huge space
hapen
WARNING: No matches found for spectrum 3706, charge 1.
terminate called after throwing an instance of 'std::bad_alloc'
what(): St9bad_alloc
William Noble
Hi Hapen,
Your error message indicates that you have run out of memory. You might try running your search on a machine with more memory or restricting your modifications. If you'd like us to look into this in more detail, we'd need to see the sample input files. Answers to your other questions are below:
INFO: Elapsed time: 5.88 s--------INFO: Created 11 peptide mods, keeping 11 with 255 or fewer aa mods # Could you explain the meaning of this line?
WARNING: SP Scoring: bin:4033 is greater than max:4000 # bin:4033 is the actual number of bins when binning peaks? How to avoid this warning in this case?
... ...WARNING: No matches found for spectrum 1178, charge 2. # This spectrum has no candidate peptides matched in database within the mass tolerance, right?
... ...INFO: Searching spectrum number 17374 (2+), search number 73570INFO: Searching spectrum number 17377 (2+), search number 73580INFO: Searching spectrum number 17382 (2+), search number 73590 # What's the search number? I noticed it's increased by step value 10
Bill
... ...
Thanks for your contributions to the proteomics community.
--
You received this message because you are subscribed to the Google Groups "crux-users" group.
To view this discussion on the web visit https://groups.google.com/d/msg/crux-users/-/VPhiYjzb_48J.
To post to this group, send email to crux-...@googlegroups.com.
To unsubscribe from this group, send email to crux-users+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/crux-users?hl=en.
Charles Grant
On Apr 12, 2012, at 9:38 AM, hapen wrote:
> WARNING: SP Scoring: bin:4033 is greater than max:4000 # bin:4033 is the actual number of bins when binning peaks? How to avoid this warning in this case?
I wanted to offer a brief addition to Bill's note. This warning is related to the calculation of the SP score. For the purposes of SP scoring, Crux assumes that no peak will have an mz larger than 4000. Peaks with higher mz are ignored in the calculation of the SP score.
An undocumented parameter, "max-mz", can be used to override this. You could try adding this parameter your parameter file. We haven't tested this, so we can't promise that it will improve your results. The line in your parameter file would look like
max-mz=4500
Charles
hapen
hapen
B. Frewen
The numbers for the modifications are as follows.
> mod <mass change>:<aa list>:<max per peptide>
The 'max per peptide' value is how many times this modification can be
applied to a peptide.
> max-mods <int>
This is the maximum number of modifications to look for on a peptide.
Suppose you have these parameters:
mod=79.9:STY:1 #phospho
mod=16:M:1 #oxidation
max-mods=3
The search will look for unmodified peptides, peptides with one phospho,
peptides with one oxidation, and peptides with one pospho and one
oxidation. Even though max-mods is 3, neither of those mods can be
applied more than once, so the defacto maximum number of mods per peptide
is 2.
Suppose a different set of parameters:
mod=79.9:STY:4 #phospho
mod=16:M:4 #oxidation
max-mods=3
Now there are many more combinations possible: no mods, one phospho, two
phosphos, three phosphos (but not four), one oxi, two oxi, three oxi, one
phospho and one oxi,...and so on.
> max-aas-modified <int>
In crux, an amino acid can have more than one modification applied to it.
This value is similar to 'max-mods', but counts each modified amino acid
as one instead of counting each modification. To use your example
> Suppose I have the following modified peptide:
> �x x �z �z
> �| | �| �|
> ANONYMOUSPEP �(x modify MN, y:NO, z:OP)
> �|| � |
> �yy � y
There are 7 total modifications on this peptide, so max-mods must be at
least 7. There are 5 modified amino acids on this peptide, so
max-aas-modified must be at least 5.
> Do the latter two include fixed mos and N/C terminal fixed/var mods?
All variable mods and fixed terminal mods are included in these counts
(parameters mod, cmod, nmod, cmod-fixed, nmod-fixed). Fixed mods with no
positional information (e.g. C=57.0) are not included.
Let me know if I can further clarify any of that.
Thanks,
Barbara
B. Frewen
I'll jump in and answer this one for Bill. There is such a filter if you
use Sp scoring. For crux search-for-matches, set compute-sp=true. By
default 500 of the top-scoring candidates will be kept and then scored by
xcorr. You can change that number to, say 200, with
max-rank-preliminary=200.
This is the default behavior for crux sequest-search. See the FAQ for
more information on the differences between the two.
http://noble.gs.washington.edu/proj/crux/crux-faq.html
Thanks,
Barbara
On Fri, 13 Apr 2012, hapen wrote:
> Hi Bill,
> Thanks a lot for your answers.I reduced the number of modifications
> (mod=11.050561:ACDEFGHIKLMNOPQRSTUVWY:4) and now it works.For the memory
> usage, if I understand it correctly, my error message came from where Crux
> try to allocate large space for candidate peptides for one spectrum in
> "MatchCollection.cpp".�If the only top-K candidates are needed, say top 500,
> how about maintaining a max-heap to keep these candidates? Then the memory
> requirement will be largely reduced.
>
> BRs/Haipeng
>
> On Friday, April 13, 2012 1:39:21 AM UTC+2, Bill Noble wrote:
>
> Hi Hapen,
>
> Your error message indicates that you have run out of memory.�
> You might try running your search on a machine with more memory
> or restricting your modifications.� If you'd like us to look
> into this in more detail, we'd need to see the sample input
> files.� Answers to your other questions are below:
>
> --
> You received this message because you are subscribed to the Google Groups
> "crux-users" group.
> To view this discussion on the web visit
> https://groups.google.com/d/msg/crux-users/-/gP7qMbezWokJ.