Question: Comparison between two different group's pathway analysis results
0
gravatar for tntntntntn
12 weeks ago by
tntntntntn0
tntntntntn0 wrote:

Hi,

I have RNA-seq dataset from two population and two conditions as recently posted in https://support.bioconductor.org/p/123895/#123921. I've successfully performed the evaluation of difference between two groups in gene expression changes thanks to you.

Then, I have done pathway analysis based on fold change between two conditions using gage with reference to KEGG pathway.

It seems that there is difference between two groups in pathway analysis results (i. e. some pathway is upregulated by condition change in one group, but not changed in the other group). I would like to evaluate which pathway is significantly different between groups. Does someone knows how to evaluate it with statistically valid method?

Thanks in advance for your help.

Regards,

Tatsuhiko

gage deseq2 pathway analysis • 147 views
ADD COMMENTlink modified 12 weeks ago by mikhael.manurung200 • written 12 weeks ago by tntntntntn0
Answer: Comparison between two different group's pathway analysis results
0
gravatar for mikhael.manurung
12 weeks ago by
Netherlands
mikhael.manurung200 wrote:

You can do either an overrepresentation analysis (ORA) or gene-set enrichment analysis (GSEA). Those two are entirely different but many seems to got it wrong.

In an overrepresentation analysis, the input is the significant genes after differential expression analysis. You can combine both up/downregulated genes in one list, however, I recommend to do the ORA separately for up and downregulated genes. You can use online tools such as gprofiler or Bioconductor package such as goseq. goseq might be a better option because it also correct your results based on gene length bias. But hey, if you want to get the results quickly, there's nothing wrong with using those web-based tools.

The second way is the GSEA. The input for this method is all your genes that have been ranked. You can rank the genes based on the fold change or moderated t values. If you are using DESeq2then I'd recommend to do lfcShrink with apeglm method. For the GSEA itsef, there are so many options out there and numerous benchmarks have been done and I can only say that each method has its own strength and weaknesses. You can also use gprofiler and then tick the Ordered query tick box (right hand side, under Options). Or you can use something like fgsea which is quite simple to use.

ADD COMMENTlink written 12 weeks ago by mikhael.manurung200

From your gprofiler link

g:GOSt performs functional enrichment analysis, also known as over-representation analysis (ORA) or gene set enrichment analysis, on input gene list.

And it seems to have just one "Run query button"

ADD REPLYlink written 12 weeks ago by swbarnes2340

Thank you for your reply. But you may misunderstand the intention of my question for the same reason as follows. It will be a great help if you may have any suggestion.

ADD REPLYlink written 12 weeks ago by tntntntntn0

If you tick the ordered query button then it will consider the input as a pre-ranked list of genes for GSEA.

ADD REPLYlink written 12 weeks ago by mikhael.manurung200

Thank you for your replay. However, you may misunderstand my question. My interest is whether there is a statistical method to compare the results of pathway analysis between two groups. I've already done pathway and GO analysis for fold changes caused by two conditions. But I would like to compare them between two groups (not condition). It will be a great help if you may have any suggestion.

Regards,

Tatsuhiko

ADD REPLYlink modified 12 weeks ago • written 12 weeks ago by tntntntntn0

I don't have any methods in mind for this.

ADD REPLYlink written 12 weeks ago by Michael Love26k

Are you looking for pathways that are regulated differently across the groups after a similar condition change? Then I guess you can test for interaction term. See this post as an example.

Hint: results(dds, type="LRT", full= ~ group + condition + group:condition, reduced= ~ group + condition)

ADD REPLYlink written 12 weeks ago by mikhael.manurung200
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 432 users visited in the last hour