How to deal with `'select()' returned 1:many mapping between keys and columns` warning when retrieving Entrez IDs using Gene Symbols in R?

Question

[melissachua90]

I want to replace the index of the dataframe (Gene Symbols) with GENE_ID:GENE_VALUE to use as a data matrix input for netgsa R package (https://cran.r-project.org/web/packages/netgsa/vignettes/netgsa.html).

First, I retrieve the Entrez IDs:

library(org.Hs.eg.db)
library(AnnotationDbi)

# gene_value is the Entrez ID
gene_value <- as.data.frame(mapIds(org.Hs.eg.db, keys=rownames(meth_df), column="ENTREZID", keytype="SYMBOL")) # gene_value is the Entrez ID

Traceback:

'select()' returned 1:many mapping between keys and columns

Then, I want to append the string ENTREZID: to the gene_value variable.

rownames(meth_df) <- paste0("ENTREZID:", gene_value)

Traceback:

Error in .rowNamesDF<-(x, value = value) : invalid 'row.names' length

Expected rownames output (example):

## [1] "ENTREZID:127550" "ENTREZID:53947"  "ENTREZID:65985"  "ENTREZID:51166" 
## [5] "ENTREZID:15"     "ENTREZID:60496"

Example data:

> dput(meth_df[1:5,1:5])
structure(list(`TCGA-2K-A9WE-01A` = c(0.611033076810465, 0.786837244239289, 
0.531054614303851, 0.711916183761331, 0.758443223998425), `TCGA-2Z-A9J1-01A` = c(0.468013052647261, 
0.386177267500376, 0.508623627469028, 0.403601275088479, 0.754642399207848
), `TCGA-2Z-A9J2-01A` = c(0.593559707995411, 0.54983504208745, 
0.535207192925841, 0.613971903755576, 0.717278085189431), `TCGA-2Z-A9J3-01A` = c(0.638211007873003, 
0.319561448644096, 0.526699541432941, 0.450002172806716, 0.736440001203422
), `TCGA-2Z-A9J5-01A` = c(0.603998109440889, 0.638039512259872, 
0.584328151056768, 0.594021097192165, 0.818583455926719)), row.names = c("A1BG", 
"A1CF", "A2BP1", "A2LD1", "A2M"), class = "data.frame")

[Basti]

I have no error with your code and your example, what is your package.version("AnnotationDbi")?

[melissachua90]

The package version for AnnotationDbi is "1.58.0"

[Basti]

You changed the post now you have no error. As some ENTREZID will be NA's for some genes, you will need to drop the genes with no ENTREZID in both your gene_value list and meth_df

[melissachua90]

It still gives the same error. The meth_df does not have missing genes (index value).

[Basti]

You changed the error you had for mapIds, now you have gene_list which is a vector containing ENTREZID for your input genes. What I would do is to merge both information, remove genes with no ENTREZID and change the rownames as expected :

meth_df=na.omit(merge(meth_df,data.frame(gene_value),by=0))
rownames(meth_df) <- paste0("ENTREZID:", meth_df$gene_value)

[melissachua90]

gene_value produced error 'select()' returned 1:many mapping between keys and columns. So I need to solve this error first before I can proceed with your recommendation.

gene_value = mapIds(org.Hs.eg.db, keys=rownames(meth_df), column="ENTREZID", keytype="SYMBOL") # gene_value is the Entrez ID

'select()' returned 1:many mapping between keys and columns

meth_df=na.omit(merge(meth_df,data.frame(gene_value),by=0))
rownames(meth_df) <- paste0("ENTREZID:", meth_df$gene_value)

Error in .rowNamesDF<-(x, value = value) : invalid 'row.names' length

[Basti]

I don't understand, you first said that you had this error : "Error in mapIds_base(x, keys, column, keytype, ..., multiVals = multiVals) : mapIds must have at least one key to match against. "

After you modified your post and it worked resulting in "'select()' returned 1:many mapping between keys and columns" warning message but it is possible to deal with it

Now in your code there is again the first error, so why is it changing ?

[melissachua90]

Sorry for the confusion. I edited my comment above. Error code is as follows:

gene_value = mapIds(org.Hs.eg.db, keys=rownames(meth_df), column="ENTREZID", keytype="SYMBOL") # gene_value is the Entrez ID

'select()' returned 1:many mapping between keys and columns

meth_df=na.omit(merge(meth_df,data.frame(gene_value),by=0))
rownames(meth_df) <- paste0("ENTREZID:", meth_df$gene_value)

Error in .rowNamesDF<-(x, value = value) : invalid 'row.names' length

Answer 1

[LChart]

If I'm using gene symbols for primary analysis, I prefer to use HGNC-provided mappings:

http://ftp.ebi.ac.uk/pub/databases/genenames/hgnc/tsv/hgnc_complete_set.txt

the 'hgnc_symbol' and 'entrez_id' columns are what you're looking for.

[melissachua90]

Would you mind to provide the code? Thanks.

[LChart]

mapping <- read.csv('hgnc_complete_set.txt', sep='\t')[,c('symbol', 'entrez_id')]
sym2ent <- sprintf('ENTREZ:%d', mapping$entrez_id)
names(sym2ent) <- mapping$symbol
rownames(thing) <- ifelse(rownames(thing) %in% names(sym2ent), sym2ent[rownames(thing)], rownames(thing))

[melissachua90]

Thanks! However, the code only edit some of the row names. For example, "ENTREZ:1", "ENTREZ:29974" and "ENTREZ:2" are the new row names from the code; but "A2BP1" and "A2LD1" are the gene symbols from the original meth_df dataframe.

Is it because some genes do not have entrez ID? If so, how should I deal with it?

rownames(meth_df) <- ifelse(rownames(meth_df) %in% names(sym2ent), sym2ent[rownames(meth_df)], rownames(meth_df))

rownames(meth_df)

[1] "ENTREZ:1" "ENTREZ:29974" "A2BP1"
"A2LD1" [5] "ENTREZ:2"

[LChart]

Is it because some genes do not have entrez ID?

Yes.

If so, how should I deal with it?

If you strictly require entrez IDs, then you have to subset to those genes:

meth_df_entrez <- meth_df[grepl('^ENTREZ', rownames(meth_df)),]

Otherwise you have to live with a combination of HGNC and Entrez.