Allowing DNA and RNA in ProForma

[douweschulte]

This issue is based on a discussion in the PSI meeting last week.

Problem

Adding DNA and RNA to ProForma would enable easier writing and reporting of these molecules for MS experiments. For example this would allow the use of these in spectral libraries stored in the mzSpecLib format (as precursors, if needed mzPAF should be extended with DNA/RNA fragmentation naming). At a high level DNA, RNA, and proteins are similar: both build up of distinct building blocks in a linear chain with a defined backbone. So writing DNA in ProForma could be done like so: ATGCTG[+34]A. The basic sequence block should be changed from amino acid to base pair and the modifications changed to a database of base pair modifications. This would still allow for all complexity of ProForma (modifications of unknown position, labile, global, chimeric, cross-linking etc). One issue that is left then is to be able to tell apart DNA/RNA sequences from peptides, especially if both are allowed to be mixed in the same definition. This should be done with some kind of tag before the now called 'peptidoform'. One proposal is to use <!DNA>, <!RNA>, and <!AA>. Examples:

<!DNA>ATGTCATCGT
<!RNA>AUGUCAUCGU
<!DNA>ATCG[Formula:OH#XL1]T//<!AA>HYTGC[#XL1]R

To my eyes <> looks like a namespacy, global, meta tag thing, a bit like an HTML tag.

This tag is unambiguously and easily parse-able, as the only current options for things starting with < are global modifications and global isotope replacements. Global modifications <[mod]@location> are easily distinguished by the use of the [ and global isotope replacements always use a number (expect for <D> but that is also easily recognised). I would propose to add this tag right before the peptidoform name tag <!DNA>(>My nice sequence 001)ATGCTAGT which means that any global modification comes before. If this tag is missing the sequence defaults to a protein/peptide which would make this addition easily backwards compatible.

Open questions

• Is there a good database for DNA/RNA modifications?
  Joshua: RNA: https://genesilico.pl/modomics/
• What are all the ambiguous base pair names we allow?
  https://www.insdc.org/submitting-standards/feature-table/#7.4.3, only thing is that in this table only T is allowed while allowing U might be easier to read and understand
• How do we handle global modifications, as currently they only allow for amino acid locations?
• Is the tag <!DNA> good enough, or can we find a better one?
• What are better names for compound peptidoform ion/peptidoform ion/peptidoform now that the whole concepts of peptides is left behind?
• What to do about double stranded DNA/RNA? Also what to do for nonstandard pairing, or modifications on the other strand?
• Do we need specific notation for backbone/base/linker localisation of modifications?
  We do not have this for amino acids, although the SMILES could provide a start for this.
• Do we need to be able to differentiate 5'-to-3' and 3'-to-5' strands?
• Having a DNA strand with some RNA bases and the other way around does happen, how do we handle this?
  Having a modification for these cases might just work.
• Is there a need for a different system for modifications of the linker?
  This might be handled by the modification database already.

Answered questions

• Is it needed to mix peptide and base pair sequences in one definition?
  Yes, apparently there are experiments conducted with DNA cross-linked to peptides. Allowing both to mix then is the most elegant way to allow all of that complexity. (For example: 10.1016/j.cell.2025.04.037)
• Is there a need to tell apart RNA and DNA or do we need a single type to represent both?
  Yes, RNA has a different sugar in the backbone. As well as the T and U base difference.
• Is there a better name for Protein in the tag, a bit smaller or better fitting in this context?
  Joshua: <!AA>

Nice figures

10.1016/j.mcpro.2024.100742

[mobiusklein]

RNA modification database: https://genesilico.pl/modomics/

[mobiusklein]

Is there a better name for Protein in the tag, a bit smaller or better fitting in this context?
<!AA> is shorter and perhaps more generic.

Another question for the world of protein-RNA crosslinking is whether or not the RNA being cross-linked with is double stranded or single stranded. If it's double stranded then we implicitly assume base-pairing rules apply, but what about when you crosslink a mutated transcription product s.t. you have a mismatch? Do we express that as a modification? Do we need a notation for differentiating single- vs double-stranded nucleotide sequences?

[douweschulte]

I like the <!AA>. For double stranded we could maybe allow an additional tag on the <!DNA double> to indicate it uses conventional pairing. But conventional pairing might be hard when we allow ambiguous bases. Additionally, we might be able to use the cross-linking notation to indicate manual base pairing: <!DNA>ACCTG//<!DNA>TGGAC. It might be fine to not require any additional explicit cross-linker, as there is none and detecting if two linear sequence stretches are both DNA and contain no cross-links to each other is doable in code. The only issue with this last syntax is that <!DNA>AGCT[#XL1]//<!DNA>TCG[+23#XL1]A//<!AA>WE[#XL1]TH is ambiguous between being a double stranded DNA or just two separate DNA strands, although I am even not sure if that matters.

[poshul]

Weighing in on this now that I've been working with adding proforma support to OpenMS:

I like the name-spacey <!AA> prefix.

The way that we've handled representing modified RNA's for NASE already bears some resemblance to ProForma (more specifically it's loose cousin of PEFF): We have an identifier line (starting with >), followed by a sequence line. Nucleic acids are represented by their modomics short code. If that code is only a single character no further formatting is necessary, if it's more than one character the whole short code gets placed inside brackets. For example AUCCUI just uses inosine as is but AUCCU[Cm] needs to be inside brackets. It should be noted that this is different from how ProForma handles mods in that the original monomer is replaced by the bracket form, rather than the bracketed form applying to the previous monomer.

This worked fine for the limited case where there were no phosphate mods and we knew ahead of time whether the riboses were deoxy-ed, and the only thing that showed up in brackets was the modomics codes. If we wanted to continue using modomics and "singlet" representation, my suggestion for extension would be to add a namespace to the modomics short codes and names, ditch the single letter "naked" representation:
the examples then become:
AUCCU[MOD:I] and AUCCU[MOD:Cm] or
AUCCU[MOD:Inosine] and AUCCU[MOD:2'-O-methylcytidine]

I'd also suggest allowing SMILES strings:
AUCCU[OC[C@@H]1[C@@H](O)[C@@H](O)[C@H]([n]2c3nc[nH]c(=O)c3nc2)O1] and AUCCU[CO[C@H]1[C@H]([n]2ccc(N)nc2=O)O[C@H](CO)[C@H]1O]

and unknown mass shifts:
AUCCU[A+0.984] and `AUCCU[C+14.0157]

A widely used alternative is the depict each nucleotide as a triplet: linker, BASE, sugar:
Ar pUr pCr pCr pUr pIr and Ar pUr pCr pCr pUr pCm or
Ar-pUr-pCr-pCr-pUr-pIr and pUr-pCr-pCr-pUr-pCm

This makes the representation of mixed RNA/DNA easier, and allows for additional arbitrary phosphate replacements, but it doesn't play nicely with Modomics since base and sugar modifications are displayed separately.

If I were implementing something from scratch, with the design goal of making it as close to ProForma as possible, and using Modomics as our ontology I'd probably do something hybrid:

• Use singlet notation by default
• Use a <!DNA> or <!RNA> tag to determine the "default" sugar for a sequence, allowing override by explicit specification of r or d at individual positions. This is a bit of an abuse of Modomics, since they don't have deoxy versions of a lot of their mods, but I think it's a worthwhile sin.
• assume phosphate linkers if not already specified (Or do we want to have another namespace tag to alter this?)
• Since Modomics has both nucleotide and nucleoside entries modomics codes should replace either the base and the sugar or the base the sugar and the linker.
  So in the hybrid form our examples above would be:

AUCCU[MOD:I] and AUCCU[MOD:Cm]

but could also be represented as:

ArUrCrCrUr[MOD:I]r and ArUrCrCrUr[MOD:Cm]r

or if you were feeling particularly verbose:

ArpUrpCrpCrpUrp[MOD:I]r and ArpUrpCrpCrpUrp[MOD:Cm]r

For something that is included in Modomics as a nucleotide:

AUCCU[MOD:pm1I] would be fine, AUCCU[MOD:pm1I]p would NOT, since the mod already contains the phosphate

For handling double stranded molecules, my druthers would be to just have the complementary sequence continue (5'-3') after the end of the first strand, separated by some non-whitespace character to denote that they are not actually linked by a phosphate bond. From there we can optionally use the ProForma crosslinking annotation to show where there is base pairing if that is not obvious from the sequence. This would then allow strands that are only partially complementary, or even have secondary structure of their own.

This was a bit of a brain dump, but I'm curious to hear any thoughts.