pairwise_termsim Throws Error if No Significant Gene Sets

[DarioS]

Error in pairwise_termsim.enrichResult(x, method = method, semData = semData: no enriched term found...

My supervisor has asked me to plot the top ten ranked sets if none are significant, but that does not seem to be possible.

[guidohooiveld]

Could you please add the upstream-code that results in this error?
Related to this, to reproduce the error, would it be possible to share your input data (as R-object)? Also a little more context would be appreciated.

[DarioS]

Yes, certainly. I thought this was by design and expected. To me, it is undesirable and rigid. I show a self-contained example.

library(enrichplot)
library(DOSE)
library(clusterProfiler)
library(msigdbr)
library("org.Hs.eg.db", character.only = TRUE)

hallmarks <- msigdbr(category = "H")
hallmarks <- split(hallmarks[["gene_symbol"]], hallmarks[["gs_name"]])
geneSet <- hallmarks[["HALLMARK_ANDROGEN_RESPONSE"]]
all <- unlist(hallmarks)
result <- enrichGO(gene = geneSet, ont = "BP", keyType = "SYMBOL", universe = all, OrgDb = "org.Hs.eg.db")
similarities <- pairwise_termsim(result)

This causes Error in pairwise_termsim.enrichResult(x, method = method, semData = semData,: no enriched term found .... Not good for plotting in R Markdown report in a loop. It necessitates the use of tryCatch but some plot would be nicer to see.

[guidohooiveld]

Thanks for providing the code. I could reproduce the error.

> result
#
# over-representation test
#
#...@organism    Homo sapiens 
#...@ontology    BP 
#...@keytype     SYMBOL 
#...@gene        chr [1:100] "ABCC4" "ABHD2" "ACSL3" "ACTN1" "ADAMTS1" "ADRM1" "AKAP12" ...
#...pvalues adjusted by 'BH' with cutoff <0.05 
#...0 enriched terms found
#...Citation
 T Wu, E Hu, S Xu, M Chen, P Guo, Z Dai, T Feng, L Zhou, W Tang, L Zhan, X Fu, S Liu, X Bo, and G Yu.
 clusterProfiler 4.0: A universal enrichment tool for interpreting omics data.
 The Innovation. 2021, 2(3):100141 

> similarities <- pairwise_termsim(result)
Error in pairwise_termsim.enrichResult(x, method = method, semData = semData,  : 
  no enriched term found...
>

On one hand I do understand that because of reasons of automation/scripting you would like that pairwise_termsim, but also the functions for subsequent generation of plots, are able to handle empty inputs (i.e. 0 enriched terms found).
On the other hand, if there are no gene sets present in the input, then pairwise_termsim can obviously not calculate any similarity values (nor can plots be made).

To be honest, I cannot think of an easy way to have enrichplot taking care of empty result files.

@huerqiang : do you have any ideas on this?

[DarioS]

The input is not empty. It is not a case of "no gene sets present in the input". There are many present.

> head(result@result)
        ID                                    Description GeneRatio  BgRatio       pvalue p.adjust qvalue
GO:0070972  protein localization to endoplasmic reticulum     5/100  24/4306 0.0001828789 0.308324 0.303
GO:0003006 developmental process involved in reproduction    17/100 302/4306 0.0004847255 0.308324 0.303
GO:0008299                isoprenoid biosynthetic process     4/100  17/4306 0.0005166236 0.308324 0.303
GO:0030521            androgen receptor signaling pathway     4/100  19/4306 0.0008118204 0.308324 0.303
GO:0000003                                   reproduction    22/100 480/4306 0.0011720568 0.308324 0.303
GO:0008610                     lipid biosynthetic process    15/100 269/4306 0.0011891493 0.308324 0.303

pairwise_termsim simply refuses to calculate anything when none of the sets have a signficant adjusted p-value.