Analyzing large networks in Cytoscape

[Mat]

Hello everyone,

I'm a beginner in Cytoscape and I want to use it for analyzing RNA seq data, by applying for example ClueGO and BinGO.

I had the idea that I could make very large protein interaction networks and subsequently load my data files onto it.

So I imported and merged the interactomes that could be downloaded by the Web Service Client (I forgot the number of nodes but about 600k edges) and loaded a data file and tried to do some analyses using these apps.

This failed however in that it got stuck at 'finalizing' for more than 24 hours. 

I'm using Cytoscape 3.2.0 and the specs of the (Windows 7 64-bit) pc are:

12 Gb Ram

Intel XEON X5650 @ 2.67 GHz

ATI FireMV 2260

Java 8 v31

So I have a few questions that I hope to see answered. 

Are these networks to big for Cytoscape to handle or are my specs just not good enough? and if the latter is the case, what specs should then work?

Finally, is the OS X or Linux version of Cytoscape better able to handle large datasets?

Regards,

Matthias

[Scooter Morris]

Hi Matthias,
    I often work with networks of that size (or much larger).  On the other hand, the Cytoscape "merge networks" function can be extremely slow.  I would suggest downloading a single interactome (there are several sample ones on the startup screen) and avoiding the merge step.  Cytoscape also gets really memory hungry for large networks, so you should watch the little memory icon in the lower right hand corner to tell you if you are running out of memory.

-- scooter

--
You received this message because you are subscribed to the Google Groups "cytoscape-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cytoscape-disc...@googlegroups.com.
To post to this group, send email to cytoscap...@googlegroups.com.
Visit this group at http://groups.google.com/group/cytoscape-discuss.
For more options, visit https://groups.google.com/d/optout.

[Ankush Sharma]

I also encountered similar problem, one way is to increase memory heap size of Java by editing .vmoption file. Hope this helps

Sincerely
Ankush sharma

[Tim Hull]

Usually the max heap size is automatically set in the Cytoscape.vmoptions file to something close to your physical memory. You may want to check to make sure that is the case - if it is lower, it could be that you're using 32-bit Cytoscape/Java.

Tim

[Mat]

Thanks everyone for responding!

@ Scooter Morris:

After trying to use the merged networks, I indeed tried to do analyses with just the IntAct interactome.

Some analyses worked, but not all, and it still took a lot of time (hours). And during these processes the memory icon stays green.

@ Ankush Sharma:

I increased the Java heap size to 10Gb, and tried to open a network (about 30K) and it is after 4 hours still stuck at finalizing with the memory icon being green most of the times.

@ Tim Hull:

I had already checked the Cytoscape.vmoptions file and it was set at 11,5 Gb max, so I don't think this is the problem.

[Scooter Morris]

On 02/19/2015 03:58 AM, Mat wrote:

Thanks everyone for responding!

@ Scooter Morris:

After trying to use the merged networks, I indeed tried to do analyses with just the IntAct interactome.

Some analyses worked, but not all, and it still took a lot of time (hours). And during these processes the memory icon stays green.

Then something else is definitely going on, I think.  Depending on the analysis you are doing, it really shouldn't take all that long, unless you are hitting some sort of limit.  What analyses are you doing on the IntAct interactome?

@ Ankush Sharma:

I increased the Java heap size to 10Gb, and tried to open a network (about 30K) and it is after 4 hours still stuck at finalizing with the memory icon being green most of the times.

@ Tim Hull:

I had already checked the Cytoscape.vmoptions file and it was set at 11,5 Gb max, so I don't think this is the problem.

This is probably too large for a 12GB machine as you really don't have 11.5GB available for Java.  10GB is probably a more reasonable size.

--
You received this message because you are subscribed to the Google Groups "cytoscape-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cytoscape-disc...@googlegroups.com.
To post to this group, send email to cytoscap...@googlegroups.com.
Visit this group at http://groups.google.com/group/cytoscape-discuss.
For more options, visit https://groups.google.com/d/optout.

-- scooter

[Mat]

Then something else is definitely going on, I think.  Depending on the analysis you are doing, it really shouldn't take all that long, unless you are hitting some sort of limit.  What analyses are you doing on the IntAct interactome?

I've run BinGO, GeneMANIA, CytoKEGG, MCODE and ClueGO of which only the first one produced results.

So you think that it's not due to the specs, since they should be able to handle it?

[Scooter Morris]

--
You received this message because you are subscribed to the Google Groups "cytoscape-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cytoscape-disc...@googlegroups.com.
To post to this group, send email to cytoscap...@googlegroups.com.
Visit this group at http://groups.google.com/group/cytoscape-discuss.
For more options, visit https://groups.google.com/d/optout.

Hi Mat,
    Do you have a session file you can share with me?  I'll take a look and see if there is anything particularly pathological about the network structure.

-- scooter

[Mat]

Hi Scooter,

Did you get my messages with the session file?

 Regards,

Matthias

[Scooter Morris]

Hi Matthias,
    Sorry, I've been pretty busy the last several days.  I was able to analyze your network using clusterMaker2's MCODE implementation, but what I did was to use "Edit->Remove Duplicate Edges" and "Edit->Remove Self-Loops" first.  When you did the network merge, you got a bunch of duplicate edges from IntAct.  These edges let you know all of the various kinds of experiments the validate that association.  However, they are not useful (and are misleading) for the purposes of any of the Cytoscape algorithms.  I wasn't able to use the UT MCODE because by default it attempts to create a side panel with all of the clusters and that takes a lot of memory.

-- scooter

[Neels]

@Scooter Morris 

Hi. I am using Cytoscape 3.4 on a Dell Tower workstation with 256 GB RAM. My cytoscape.vmoptions file shows 

-Xms2048M

-Xmx248690M means 248.69 GB heap memory.

Now, I am stuck at the merge network option for almost 24 hours. I really need to do this since I have to create a large network for my organism of interest as it is not present in Cytoscape. Any suggestions??