Errors with 12/13 release?
Danielle Lemay
Is the current error that you're chasing with the 12/13/09 release
strictly with the Windows version? I've tried my files with the
cross-platform version on Linux and the online web service and am not
able to get either to work. Error messages below.
Thanks,
Danielle
(Danielle Lemay dglemay at ucdavis dot edu)
happygrad:~/work/ext_source/
--species Hs --input ../../nurit_projects/GO_input --denom
../../nurit_projects/GO_denom --output ../../nurit_projects/GO_output
Importing previously parsed version of GO
* * * * * * * * * * * * * * * * * * * * GO categories imported and
nested in 21 seconds
Performing mappfinder analysis on go_EntrezGene_union_lactosome.txt
Importing previously parsed version of GO
* * * * * * * * * * * * * * * * * * * * GO categories imported and
nested in 29 seconds
Traceback (most recent call last):
File "GO_Elite.py", line 35, in <module>
import mappfinder; reload(mappfinder)
File "/home/dglemay/work/ext_source/GO-Elite_120beta/mappfinder.py",
line 31, in <module>
import GO_Elite
File "/home/dglemay/work/ext_source/GO-Elite_120beta/GO_Elite.py",
line 1722, in <module>
if len(sys.argv[1:])>1 and '-' in command_args: commandLineRun()
File "/home/dglemay/work/ext_source/GO-Elite_120beta/GO_Elite.py",
line 1714, in commandLineRun
root = parent; runGOElite(mod)
File "/home/dglemay/work/ext_source/GO-Elite_120beta/GO_Elite.py",
line 1017, in runGOElite
dir_list = read_directory(import_dir) #send a sub_directory to a
function to identify all files in a directory
File "/home/dglemay/work/ext_source/GO-Elite_120beta/GO_Elite.py",
line 46, in read_directory
dir_list = unique.read_directory(sub_dir); dir_list2 = []
File "/home/dglemay/work/ext_source/GO-Elite_120beta/unique.py",
line 66, in read_directory
if (':' in sub_dir) or ('/Users/' == sub_dir[:7]) or ('/Volumes/'
in sub_dir) or ('Linux' in platform.system()): dir_list =
os.listdir(sub_dir) ### Thus, the whole path is provided already
OSError: [Errno 2] No such file or directory:
'../../nurit_projects/GO_output/GO-Elite_results/CompleteResults/ORA'
happygrad:~/work/ext_source/GO-Elite_120beta>
Results with online web service:
Traceback (most recent call last):
File "GO_Elite.py", line 35, in ?
import mappfinder; reload(mappfinder)
File "/usr/local/projects/cytoscape/GO-Elite_120beta/mappfinder.py",
line 31, in ?
import GO_Elite
File "/usr/local/projects/cytoscape/GO-Elite_120beta/GO_Elite.py",
line 1722, in ?
if len(sys.argv[1:])>1 and '-' in command_args: commandLineRun()
File "/usr/local/projects/cytoscape/GO-Elite_120beta/GO_Elite.py",
line 1714, in commandLineRun
root = parent; runGOElite(mod)
File "/usr/local/projects/cytoscape/GO-Elite_120beta/GO_Elite.py",
line 1017, in runGOElite
dir_list = read_directory(import_dir) #send a sub_directory to a
function to identify all files in a directory
File "/usr/local/projects/cytoscape/GO-Elite_120beta/GO_Elite.py",
line 46, in read_directory
dir_list = unique.read_directory(sub_dir); dir_list2 = []
File "/usr/local/projects/cytoscape/GO-Elite_120beta/unique.py",
line 66, in read_directory
if (':' in sub_dir) or ('/Users/' == sub_dir[:7]) or ('/Volumes/'
in sub_dir) or ('Linux' in platform.system()): dir_list =
os.listdir(sub_dir) ### Thus, the whole path is provided already
OSError: [Errno 2] No such file or directory:
'/usr/local/jakarta-tomcat-5_0_30/webapps/ROOT/app1261407470268/GO-Elite_results/CompleteResults/ORA'
Nathan Salomonis
Thank you so much for your quick results. I hadn't seen any of these
specific issues until now (the other issues were unclear and not
specific like this). With these in hand we will have a better idea of
how to correct these ASAP. More soon. If you have a dataset (input and
denom files) you want to send me directly to work with it might help,
but if not, don't worry. For linux, it looks like you used the command
line options, correct?
Best,
Nathan
> --You received this message because you are subscribed to the Google Groups
> "GO-Elite" group.
> To post to this group, send email to go-e...@googlegroups.com.
> To unsubscribe from this group, send email to
> go-elite+u...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/go-elite?hl=en.
>
Andreia Fonseca
I am getting the same errors as Daniellle, with linux and with the online version as well with windows.
Andreia
--
Nathan Salomonis
The problem has been that I have not been able to reproduce the
problem on any computers, which means that it is probably a specific
dataset issue. The reason I'm not seeing the actual bug from the error
reports is that I was not handling exceptions very well where the OBO
files are imported and MAPPFinder is run. That is now fixed. I just
uploaded a version that should show me exactly where the problem is on
the webservice version. Can you guys use your input and denominator
files on this website and report back the errors you get? For
permutations, select 0, since this should run in only 3 minutes.
Best,
Nathan
Nathan Salomonis
GO-Elite needs to have the extension ".txt" for input and for
denominator files, that's why it looks like the program is crashing. I
just ran your files without ".txt" at the end and the program crashed
with the same error you sent, but when I added ".txt" the program
worked fine. I will make sure that the version posted in the next few
days warns you about this rather than crashes.
It looks like this problem is different than Danielle's, since I can
see that her text file has the extension ".txt".
Best,
Nathan
On Mon, Dec 21, 2009 at 11:11 AM, Andreia Fonseca
Andreia Fonseca
thanks, I am feeling stupid.well it works fine! one question the Adjusted P is the Benjamini-Hochberg false discovery rate p-value?
Kind regards,
Andreia
Nathan Salomonis
Actually, this is still a real bug, so thanks for helping to find it.
The Adjusted P is the Benjamini-Hochberg false discovery rate p-value
calculated in the same way as in the R package p.adjust.
Best,
Nathan
On Mon, Dec 21, 2009 at 1:18 PM, Andreia Fonseca
Andreia Fonseca
In the page after_running elite you advise to "Select the row containing the column headers (e.g. row 18 for the local file) and then select the menu item “Data”, “Filter”, and “Autofilter”. A series of pull-down menus appear above each field in this row. We recommend filtering based on the “Number Changed” (>2), “Z Score”(>1.96), and “PermuteP”(<0.05). Performing this filter or other user defined filters will reduce the set of results from hundreds to a handful of pathways that are over-represented in your analysis"
the permuteP is not corrected for multiple testing right? In my case I have PermuteP<0.05 but the AdjustP is 0.5 so, these pathway are not really over-represented right?
Kind regards,
Andreia
Nathan Salomonis
You bring up some important points. The section of the tutorial and
help you are referring to does suggest to filter based on the PermuteP
and not the Adjusted FDR p value. You can do either, however, in the
pruning step, currently, you don't have the option to filter based on
the AdjP. We can add this option to the pruning step. In general, I
recommend looking at the permute and adjusted p values and seeing
whether the adjusted p value can be used to report a sufficient set of
meaningful terms or if you need to filter based on the adjusted p
value. The same goes for regular gene expression analyses and using
adjusted p values. If you have prior evidence that certain genes are
changed, but they do not meet the predetermined AdjP in your
microarray dataset then you may need to use the RawP and more
carefully look at the variance in each focused result.
Ideally, you would just use the Adjusted p at a threshold that you
choose (e.g., 0.05). However, as you mentioned, in many analyses, this
adjusted p-value is two stringent, especially when only a smaller set
of genes are regulated. Also, the permutation p value restricts the
significance, when a smaller number of permutations are run. The order
of significance for the PermuteP and thus the AdjP is dependent on the
number of permutations run (the more zeros in the number of
permutations the more decimal places in the resulting p values).
However, it is not always reasonable to run 10000 or more
permutations, since this will significantly increase the run time (5X
with 2000 permutations). In the end, use the AdjP if it will report
terms which you believe are meaningful looking back at the data or use
more stringent PermuteP filtering (e.g., >0.01) if you are worried
that a non-adjusted p is allowing for too much noise. A statistician
will probably have more to say about this if you consult them, since
this is a common issue in the analysis of any large scale data.
Best,
Nathan