sam-tags: MM field value pattern does not allow ambiguity codes

[zaeleus]

This is in regard to Sequence Alignment/Map Optional Fields Specification (2022-08-17).

The base modifications (MM) field allows modifications to be either short codes or an ChEBI ID. Short codes are constrained to [a-z]+ (i.e., lowercase letters) but the table of "standard common types" lists ambiguity codes that do not match this (i.e., uppercase letters).

Unmodified base	Code	Abbreviation	Name	ChEBI
C	C		Ambiguity code; any C mod
T	T		Ambiguity code; any T mod
U	U		Ambiguity code; any U mod
A	A		Ambiguity code; any A mod
G	G		Ambiguity code; any G mod
N	N		Ambiguity code; any mod

[jkbonfield]

The short codes are modified bases, so "m" and "h" being 5mC and 5hmC. It doesn't make any sense to have a base modification from nucleotide to ambiguity code, so I'm not sure I follow this.

We don't support ambiguity codes in the unmodified base component, so we couldn't do MM:Z:Y+h,4; for example as it wouldn't may sense. "N" covers this case anyway with the different counting regime.

[zaeleus]

I'm referring to the Code column of the standard common types table under the MM description. It defines codes that are uppercased, but the MM field pattern does not allow it: MM:Z:([ACGTUN][-+]([a-z]+|[0-9]+)[.?]?(,[0-9]+)*;)*. I referred to the short code portion as [a-z]+ originally.

The description for ML gives an example of using an ambiguous modification:

For example MM:Z:C+C,10; ML:B:C,229 indicates a C call with a probability of 90% of having some form of unspecified modification."

See that it uses C as the modification code, which does not match ([a-z]+|[0-9]+).

[jkbonfield]

Oh wow I'd totally forgotten about that!

Yes, the regexp should be ([a-zACGTUN]+|[0-9]+) for the code portion. Good spot. Thanks :)