Name of blocking variable not correctly recognized about differentialabundance HOT 7 OPEN

This bug is due to this line of code:

It probably needs to be changed to:

it.blocking = it.blocking.replaceAll('^NA$', '')

from differentialabundance.

The comment directly before this code block reads:

// Split the contrasts up so we can run differential analyses and
// downstream plots separately.
// Replace NA strings that might have snuck into the blocking column

Why would this code block try to replace NA sub-strings in column headers of the samplesheet.csv file? Maybe you apply it to samplesheet.csv, too?

from differentialabundance.

The issue is that the current code replaces any "NA" in the string. Mine only replaces it if it's the whole string. However, come to think of it, IDK why it's replacing it in the variable name anyway, and whether that's desired behavior

from differentialabundance.

I assume (carefully) that .splitCsv might add NAs when it finds an empty column; is that correct, @pinin4fjords?

Edit: No it doesn't, at least not in a little test I ran. Why might NAs sneak in? 🤔

from differentialabundance.

Thanks for the bug report!

This was definitely done in response to a bug, possibly people using NA in the input contrasts files to indicate missing values. So I don't want to remove this entirely.

@BEFH - could you PR your regex fix please, since it's so concise? Please add a changelog entry in the same style as the others there when you do so.

We should also document the special meaning of 'NA'.

from differentialabundance.

Hey, checking for and dealing with NA entries in contrasts.csv seems reasonable. But, replacing any pattern match of 'NA' with the empty string will destroy meaningful entries like 'RNA_CONCENTRATION' or 'ANALYSIS_OUTCOME' and so on. Would it be possible to replace 'NA' with the empty string only if the complete entry is 'NA' (in perl ^NA$) but otherwise don't replace NAs?

from differentialabundance.

Yes, that's the fix proposed by @BEFH

from differentialabundance.