J-Express forum
December 20, 2014, 03:18:54 PM *
Welcome, Guest. Please login or register.

Login with username, password and session length
News: J-Express 2011 released. Download from www.molmine.com
 
   Home   Help Search Login Register  
Pages: [1]
  Print  
Author Topic: Jython scripting challenge: Automatic gene group multi-fss  (Read 7811 times)
Trond Hellem Bø
Developer
Newbie
****
Posts: 12


View Profile
« on: October 20, 2005, 12:41:55 PM »

I have something of a Jython scripting challenge. I'm wondering whether it would be realtively easy to do it with Jython scripting in J-Express or if I should make a separate system for this. I'm going to do this one way or another, either in J-Express or by making a separate program.

Here is the problem:

Suppose I have a set of genes that I want to test against several clinical variables. I "know" from literature that the genes in my gene set are supposed to be significantly associated with at least one of my clinical variables. So what I would like to do is to test all genes in the gene set (defined as a gene group in J-Express) against all clinical variables (assume for simplicity they are defined as sample 2-class groups in J-express) using the Wilcoxon rank sum z-approximation (available under FSS in J-Express, or alternatively SAM which is available in J-Express 2.7), and store the results in a data sheet that can be saved to a file or copied into Excel (if your'e a Windows-user). Instead of z-scores I want P-values! (not adjusted for multiple testing, I'll worry about that later) These P-values should be based on permutations (in know there's a option under "make subset" in J-Express). For instance I would like to do 100-1000 rounds of permutations to estimate the z-score distribution (I know it's called z-score, but it's not exactly N(0,1) distributed), and use this to determine the P-values.

I guess such a test would be of interest many people. I often get questions about "where is my gene?" or "how did gene X do? How about gene Y?" when I send off analysis results. Is it doable, and is it a lot of work? I can have such a system ready in approximately a day outside J-Express since I have some code already that I can recycle for this purpose.

Cheers
Trond :-)
Logged
Bjarte Dysvik
Administrator
Full Member
*****
Posts: 116


View Profile Email
« Reply #1 on: October 20, 2005, 03:12:08 PM »

This sounds like a thing that would fit fine inside a script.. However, if you already have a lot of code outside J-Express, you could put it in a jar-file and put the jar-file in the lib folder and use classes directly from a script. Often there are things that should be repeated many times and instead of returning to the script interpreter in each iteration, you can simply create a java class that does this and use it from the script interface instead.

The problem seems like something we should already have so let's figure a way to implement this.

Bjarte.
« Last Edit: October 20, 2005, 04:24:30 PM by Bjarte Dysvik » Logged
Trond Hellem Bø
Developer
Newbie
****
Posts: 12


View Profile
« Reply #2 on: October 20, 2005, 03:52:40 PM »

Well, my code isn't exactly ready for use. I'll have to rewrite some of it to fit this problem, so it isn't exactly ready to put in the lib-directory. After rewriting it I might as well make a separate system, as the rest would be little work. We could try making a J-Express version once I've finished my stand-alone version. Shouldn't be much work if we can use the java objects from a script.

Trond :-)
Logged
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.13 | SMF © 2006-2008, Simple Machines LLC Valid XHTML 1.0! Valid CSS!