Question: Analysis of pathways using edgeR
0
18 months ago by
fawazfebin30
fawazfebin30 wrote:

Hi

I was analysing public RNA-Seq datasets from GEO for arriving at the most affected pathways /GO terms using the edgeR program. My objective was to determine whether Ca signalling pathway had a considerable effect in the corresponding phenotype. I did a DE analysis with edgeR and arrived at result in which the Ca signalling pathway was one among the top (topKEGG) hundred pathways :

(No.80)                         Pathway                       N   Up  Down       P.Up                 P.Down

path:hsa04020             Calcium signaling pathway  79   8    5            3.534894e-02          3.626250e-01      (at 2hr)

(No.72)                       Pathway                        N   Up   Down       P.Up               P.Down

path:hsa04020             Calcium signaling pathway  79   14    12            1.795998e-02     3.278448e-01       (at 4hr)

What can I conclude on this pathway regarding its effect ?  Thanks in advance.

modified 18 months ago by Gordon Smyth38k • written 18 months ago by fawazfebin30
Answer: Analysis of pathways using edgeR
3
18 months ago by
Gordon Smyth38k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth38k wrote:

Well, the analysis shows that your up-regulated genes are moderately enriched for Ca signally pathway genes (P = 0.035 at 2 hours and P = 0.018 at 4 hours).

Down-regulated genes are not enriched (P = 0.36 at 2 hours and P = 0.33 at 4 hours).

Further interpretation is a biological question, and is up to you!

If you are only interested in the KEGG Calcium Signalling pathway, you might consider doing a fry test for this pathway:

> KEGG <- getGeneKEGGLinks(species.KEGG = "hsa")
> HSA0420 <- KEGG$GeneID[ KEGG$PathwayID=="path:hsa04020" ]
> fry(y, index=HSA0420, design=design, contrast=yourcontrast)

Thank you for the valuable guidance ,Sir. Is 'y' the DGElist object in the last command of fry test ? I get the following error while running it :

> fry(y.DEX, index=HSA0420, design=design.DEX, contrast="ResistantvsControl.2hr")
Error in .zscoreDGE(y = y, design = design, contrast = contrast) :
contrast doesn't match any column of design
1

Well, the error message is self-explanatory, is it not?

Yes, 'y' is the DGEList that you used for your edgeR analysis.

'contrast' is just the same contrast that you used for your edgeR DE analysis. It has the same meaning in fry() as it has in glmLRT() or glmQLFTest(). You must already know how to specify the contrast in edgeR, otherwise you couldn't have done a DE analysis in the first place.

Yes Sir, I had used the same DGElist object and the same contrast used in DE analysis here..but couldnt understand why the error came still..Thanks
1

Type

colnames(design.DEX)

to see what the columns of the design matrix are.

You must have changed at least one of the design matrix or the contrast when running fry.

> colnames(design.DEX)
[1] "control"       "resistant.2hr" "resistant.4hr"

> DEX.contrasts
Contrasts
Levels          resistant.2hrvscontrol   resistant.4hrvscontrol
control                           -1                     -1
resistant.2hr                      1                      0
resistant.4hr                      0                      1

>  fry(y.DEX, index=HSA0420, design=design.DEX, contrast="resistant.2hrvscontrol")
Error in .zscoreDGE(y = y, design = design, contrast = contrast) :
contrast doesn't match any column of design

I am confused at the error. What could be wrong here? Thanks in advance.

1

​Perhaps you intended to do this:

fry(y.DEX, index=HSA0420, design=design.DEX,
contrast=DEX.contrasts[,"resistant.2hrvscontrol"])

Yes, Sir. Great thanks.

> HSA0420 <- KEGG$GeneID[ KEGG$PathwayID=="path:hsa04020" ]
> fry(y.DEX, index=HSA0420, design=design.DEX,
+     contrast=DEX.contrasts[,"resistant.2hrvscontrol"])
NGenes Direction PValue PValue.Mixed
set1      0        Up    NaN            0

Can the output of the above fry function be interpreted as the calcium signalling pathway is over represented ? Thanks in advance.

1

Obviously not. fry is telling you that there are no genes in the set, and so it's returning a nonsense p-value of NaN. Are you sure that the row names of y.DEX are Entrez IDs?