Integration of transcriptomics and proteomics: difficult matching names

Question

[ntsopoul]

Hi to all,

I want to integrate proteomics data and transcriptomic data but I have problems finding common identifiers.

For proteomics, I have Uniprot-ID and Gene_name(by unirprot); for mRNA-seq, I have gene_id (mm9.refGene). I looked into the mRNA-seq .gtf file and there is only one transcript ID that I could use as an alternative.

To illustrate the problem:

Nup42 (UniProt Gene_name) does not match (Nupl2 Gene_id).

Is there a way to convert the Gene_id from mRNA-seq to Uniprot Gene_name?

What are the best common identifiers for mRNA and protein data?

[GenoMax]

This table from Jackson labs that correlates many mouse identifiers with external databases will help. Look at the headers of the columns to find the various databases.

https://www.informatics.jax.org/downloads/reports/MRK_Sequence.rpt

[ntsopoul]

looks good but I cannot find either Nupl2 nor Nup42 in the list!

[GenoMax]

I can see this in the list

GI:2387631  Nup42   O   Gene    nucleoporin 42  10.72   5   24369961    24389011    +   AA867018|AB067574|AI426644|AK077066|AK078478|AK183925|AK194837|AK207606|AK216216|AV116043|BB396657|BC033270|BQ442924    NM_001346582|NM_153092|XM_030254373 ENSMUST00000049887|ENSMUST00000115101|ENSMUST00000124150|ENSMUST00000147392 Q8CIC2  A0A0R4J1K6|E9QL43   ENSMUSP00000062766|ENSMUSP00000110753   NP_001333511|NP_694732|XP_030110233     protein coding gene

UniProt also has a id-mapping tool that you may find useful: https://www.uniprot.org/id-mapping/

[ntsopoul]

you are right! Excel just did not find it...

[sansan_96]

Hello,

This conversion tool has been useful to me, maybe it can help you:

https://biit.cs.ut.ee/gprofiler/convert

[ATpoint]

For such analysis I always first translate everything to Ensembl Gene IDs (biomaRt is a good help here) and then I do the matching with this. These different identifiers and names are a pest, and imo gene ID is the only real universal constant (at least within the same gene annotation version).

Answer 1

[ntsopoul]

I found a good solution but it is really cumbersome. I use UniProt.ws to derive unique gene names (readable names) and also ambiguous names. I then collapse the columns of unique and ambiguous names into one column, so I have all possible combinations per gene/protein. Later, when I want to integrate with another data set, I separate each possible name into one row so any potential match will work.

data <-  "data/to/protoemics/"

#find gene names corresponding to uniprot ID
library(UniProt.ws)
new_names <- mapUniProt("UniProtKB_AC-ID", "Gene_Name", query =data$UniprotAccession) #unique gene names (Uniprot)
new_names2 <- mapUniProt("UniProtKB_AC-ID", "UniProtKB", query =data$UniprotAccession) #UniProtKB contains a lot of info and multiple protein names
new_names3 <- merge(new_names, new_names2[c("From", "Length","Gene.Names")], by.x = "From", by.y="From")
proteomics_named <- merge(new_names3, data, by.x="From", by.y="UniprotAccession")

# Combining two columns with a space (Gene.Names contains only ambiguous gene names and needs to be combined with the unique gene names columns)
proteomics_named <- proteomics_named %>%
  mutate(Gene.Names = paste(Gene.Names, To, sep = " "))

#now duplicated gene names might occur in the column use this function to clean up
remove_duplicates <- function(text) {
  words <- strsplit(text, " ")[[1]]  # Split the string into words
  unique_words <- unique(words)      # Remove duplicate words
  paste(unique_words, collapse = " ")  # Collapse into a single string
}

# Applying the function to the 'Text' column to remove duplicated names
proteomics_named <- proteomics_named %>%
  mutate(Gene.Names = sapply(Gene.Names, remove_duplicates))

#now separate all gene.names (multiple gene.names per row present) for better merging
proteomics_named <- proteomics_named %>%
  separate_rows(Gene.Names, sep = " ")