Trouble with Orthomcl Clusters

[jflowers002]

I recently did analysis on over 100 genomes within the same phylum using Orthomcl. Sadly, once I finally got the results a couple of weeks later, I discovered that several clusters appeared to have the same function. In fact, in one instance, there were 11 clusters that were likely fructose-2,6-bisphosphatase. I played around with the inflation value a little and found that by it resulted in clusters that appeared to be too mixed in terms of function yet there were still several repeat function clusters. Clearly, this could be a result of bad annotation, but I wanted to see if anyone has had similar problems with Orthomcl cluster prediction.

Thanks!

[jflowers002]

Well, I will answer my own question. One inherent issue that i believe caused this trouble was my blast parameters. I had 100 genomes and I only allowed 250 hits since I was concerned about diskspace and time. Orhtomcl acutally recommends not limiting this (https://docs.google.com/document/d/1...xqDAMjyP_w/pub).

I figured that 2.5 orthologs in each genome for the same gene was enough, but maybe not. I am currently rerunning the process with a larger allowed hits and hopefully this will fix it.

[bckirkup]

I'm interested to see how this has worked out for you now. Did it solve the problem?

[rhinoceros]

Depending on your research question, it might be a better idea to cluster your proteins based on hmmer searches against Pfam. This is something that can take minutes on a 4 core laptop (with say 100 bacterial genomes) vs. days on a 100 core cluster (when going with something blast-based).