collapsePathways function throwing error

Question

[Pac314]

I am trying to perform GSEA on DESeq2 results by ranking the t stat and successfully ran fgsea to extract a list of enriched pathways. However, when I run the collapsePathway function, I receive this error message:

Error in preparePathwaysAndStats(pathways, stats, minSize, maxSize, gseaParam,  : 
  Not all stats values are finite numbers

My code:

res.fgsea <- fgsea(pathways = gmt.file, stats=ranked_genes, 
                       minSize=15, maxSize=500)

cp.res <- collapsePathways(fgseaRes=res.fgsea[order(pval)][padj < 0.05], 
                                           pathways = gmt.file, 
                            stats=ranked_genes)

Please could you help me with this issue.

[Tianxia]

I encountered the same issue while I am sure the data type of stats are finite numbers. Is there a solution to this issue now? (PS: I was analyzing GSE48301 using hugene10sttranscriptcluster.db and GO file (c2.all.v2023.2.Hs.symbols.gmt.txt))

Answer 1

[H_y]

Hello, I've had the same error. In my case, the names of ranked_genes contain NA values.  I recommend that you check this. table(is.na(names(ranked_genes)))

[txema.heredia]

I just had exactly this issue. Thanks!

Answer 2

[Trivas]

I think your issue is that you're sending ALL pathways to collapsePathways and not those present in your reduced fgseaRes object.

[Pac314]

Ok, should I filter the pathways gmt file to include only those in fgseaRes?

[alserg]

I don't think so, collapsePathways filters the pathway list in the begining to contain only the pathways present in the filtered fgseaRes table.

Answer 3

[alserg]

Hi, as the error says, it means that some of your values in ranked_genes seem to be non-finite numbers (such as NAs or Infs). Can you check that all(is.finite(ranked_genes)) is true? If not, you should do something with these genes, e.g. remove them from your vector.

[Pac314]

Running all(is.finite(ranked_genes)) returns TRUE so I am not sure why it is saying that not all the stats values are finite.

[alserg]

Can you please submit an issue with a reproducible example at https://github.com/ctlab/fgsea/issues ?