Valid row/colnames after reading mzTab

[sneumann]

Unusual bug report: when reading mzTab as in https://github.com/lgatto/MSnbase/blob/master/tests/testthat/test_MzTab.R#L9, I am getting valid names in e.g.

> colnames(xx@SmallMolecules)
 [1] "SML_ID"                                "SMF_ID_REFS"                          
 [3] "database_identifier"                   "chemical_formula"                     
...
[17] "abundance_assay.4."                    "abundance_assay.5."                   
[19] "abundance_assay.6."                    "abundance_study_variable.1."          
[21] "abundance_variation_study_variable.1." "abundance_study_variable.2."          
[23] "abundance_variation_study_variable.2." "opt_global_Progenesis_identifier"     

> length(xx@SmallMolecules[,"abundance_assay.3."])  ## Works with [-operator
[1] 55

> length(xx@SmallMolecules$abundance_assay.3.)      ## Also works with $-operator
[1] 55

Which means I can't use 'em as-is when writing mzTab back out to file.
I'd have to do some string-foo and search&replace numbers enclosed in '.',
and even that could go wrong under some conditions.

OTOH the metadata names are fine (actually: illegal in R):

> names(xx@Metadata)[1:6]
[1] "mzTab-ID"                   "software[1]"                "ms_run[1]-location"        
[4] "ms_run[1]-scan_polarity[1]" "ms_run[1]-format"           "ms_run[1]-id_format"  ...

> xx@Metadata$'assays[1]'   ## Can't access a list element by illegal name with $-operator:
NULL

> xx@Metadata[['assay[1]']] ## Need to access a list element by [[-operator:
[1] "3injections_inj1_POS"

=> I'd like to propose to accept and use the illegal row/col names when it comes to mzTab,
and deal with it in the downstream code where that happens.

Objections ?

Yours, Steffen

[lgatto]

The naming in mzTab is looking for issues - that [1]... really?

The reason why the names in the metadata are the way they are (illegal) is because they don't stem from a dataframe, are casted to a list and reshaped, thus preserving their illegality.

How would you go about to use illegal column names?

[sneumann]

As shown above, the [ operator works, it needs quoting anyway: xx@Metadata[['assay[1]']]
What doesn't work is the $ operator, even with quoting. In the current MSnbase I couldn't find an example where the "fixed" colnames are required creating some MSnbase objects.
What do the assay names look like if I load an example file ? Or are they just overwritten
with the correct names from the MTD section ? Yours, Steffen

[lgatto]

You would need to use backticks for the irregular names:

> xx@Metadata$`assay[1]`
[1] "3injections_inj1_POS"

The reason the colnames are "fixed" for the SmallMolecules (as well as Proteins, Peptides, PSMs) is that these are tabular data created by read.delim and adjusted by make.names:

> make.names("assay[1]")
[1] "assay.1."

If would be possible to unset this by setting check.names = FALSE, which would return:

> xx <- MzTab(fl) ## updated code
> xx
Object of class "MzTab".
 Description:
 Mode: 
 Type: 
 Available data: SmallMolecules  
> names(xx@SmallMolecules)
 [1] "SML_ID"                               
 [2] "SMF_ID_REFS"                          
 [3] "database_identifier"                  
 [4] "chemical_formula"                     
 [5] "smiles"                               
 [6] "inchi"                                
 [7] "chemical_name"                        
 [8] "uri"                                  
 [9] "theoretical_neutral_mass"             
[10] "adduct_ions"                          
[11] "reliability"                          
[12] "best_id_confidence_measure"           
[13] "best_id_confidence_value"             
[14] "abundance_assay[1]"                   
[15] "abundance_assay[2]"                   
[16] "abundance_assay[3]"                   
[17] "abundance_assay[4]"                   
[18] "abundance_assay[5]"                   
[19] "abundance_assay[6]"                   
[20] "abundance_study_variable[1]"          
[21] "abundance_variation_study_variable[1]"
[22] "abundance_study_variable[2]"          
[23] "abundance_variation_study_variable[2]"
[24] "opt_global_Progenesis_identifier"     

As above, the $ operator would need to be used with backticks.

Is that what you are looking for to make it easier to write back MzTab data?