Gene Ontology enrichment

Question

[Czh3]

Can anyone tell me what tools should I use in gene ontology analysis - David or Ontologizer.jar?

When I run:

 java -jar ~/tools/Ontologizer.jar \
    -a ~/database/gene_association.goa_human.gz \
    -g ~/database/go.obo    \
    -n -o GO -d 0.1 \
    -m Bonferroni   \
    -s 1vs2_down.DEGs.genelist.txt \
    -c Term-For-Term    \
    -p ~/RNA_seq/KDM3B/pipe_out/GFold/human.all.gene

It is different from what David given.

Can anyone help me, why the results are different?

Any suggestion is appreciated~

[Sam]

David hasn't been updated since 2008 or 2010 I think. And the Gene ontology has been constantly updated. Therefore most likely the difference is due to the GO version. You can also try GOrilla, think they update their database more often

[Czh3]

Thanks Sam,

Yes, I have tried GOrilla, It's very useful. I found the results are the same as David, And GOrilla gives a nice DAG picture. But, I have a lot samples to run GO, SO I need a command line version of GO tools. "Ontologizer" is a command line tools to help do the GO enrichment and draw the DAG pic. But I don't know where I am wrong when running Ontologizer. Do you have any suggestion of the arguments of the Ontologizer?

[Sam]

When you said it is different, what do you mean by different? Does the terms differ or only the p-value? I was actually surprise that you have the same result for David and GOrilla as the GO was rather different. As for Ontologizer, I have never used it, so sadly I don't really know what is the best parameter

[Czh3]

Thanks a lot. What i mean 'same' are the significant GO terms which two tools given under the default p value cutoff.

[Sam]

Then how about you look into the rank of the GO terms? Are they similar? Because from the manual, Ontologizer uses bonferroni correction and I think the other two uses FDR which is slightly less stringent? So:

1. See if the rank of the GO terms are similar

2. See if the significance from Ontologizer is a subset of those from David and GOrilla?

[Gary]

Based on my experience, it is normal to get different significant GO terms when I use different tools. One possible explanation is that DAVID and GOrilla used similar statistical methods, but Ontologizer used a very different one. DAVID is still useful even its last updated time is in 2009. There are many other factors could make different results, e.g. database, multiple testing correction (both have been suggested by Sam) and gene nomenclature system. You may be not wrong when running Ontologizer, but just because that different tools use different strategies to identify enriched GO terms. Here is a useful paper for your reference. It lists a lot of gene ontology enrichment tools in the table1 and their detail in the supplement table 1 – table 3. I hope this is helpful.

Khatri P, Sirota M, Butte AJ. 2012. Ten years of pathway analysis: current approaches and outstanding challenges. Plos Computational Biology 8.

[poisonAlien]

How about goseq?

Answer 1

[Guangchuang Yu]

You can try my package, clusterProfiler.