Sequence Edits

A peptide from a sequence variant differs from the reference protein: a single substitution, several adjacent residues replaced, residues deleted or inserted. The peptidoform records only the observed sequence, and ProForma has no notation for a change relative to a reference, so the change itself would otherwise not be recorded anywhere. QPX records it as a sequence edit: a span of the reference protein and what replaces it.

The sequence_edit struct

sequence_edit: struct{
    start:       int32,          -- first reference residue of the span (1-based)
    end:         int32,          -- last reference residue of the span (inclusive)
    reference:   string,         -- the reference residues of start..end ("" for an insertion)
    alternative: string,         -- what replaces them ("" for a deletion)
    name:        string, null,   -- e.g. HGVS protein notation "p.Arg175His"
    accession:   string, null    -- e.g. a dbSNP, COSMIC or UniProt variant identifier
}

Coordinates are the reference protein's, 1-based and inclusive, so reference holds exactly end - start + 1 residues. An insertion has an empty reference and end = start - 1: it goes before residue start.

Insertion coordinates differ from VCF and HGVS

VCF and gnomAD anchor an insertion to the preceding reference residue (REF=P, ALT=PAA at 49). A sequence edit has no anchor: the insertion is start=50, end=49, reference="", alternative="AA". Converters from anchored notations must drop the anchor residue and add 1 to the position.

Change	start	end	reference	alternative	name
Substitution	175	175	R	H	p.Arg175His
Two adjacent residues replaced	10	11	KE	QK	p.Lys10_Glu11delinsGlnLys
Deletion	30	30	G	``	p.Gly30del
Insertion before residue 50	50	49	``	AA	p.Pro49_Leu50insAlaAla

Several edits on one sequence are several entries, sorted by start and not overlapping: each edit starts after the previous one ends. An edit covers reference residues start..end, and an insertion before residue s sits between s - 1 and s. Two edits of the same residue (R175H and R175C) cannot describe one sequence; alternative variants are separate proteoforms or separate peptide mappings.

proteoform.edits is filled for variant members only. An isoform or chain is also a change from the canonical sequence (a splice isoform is often a multi-residue delins), but these members are defined by their own entry (an isoform accession, a chain feature with its span) and not by a list of edits. A producer that needs the edits of an isoform records them in the pepmap of its peptides.

Where edits are stored

View	Field	Struct	Meaning
pepmap	sequence_edits	peptide_sequence_edit (a sequence_edit with protein_accession)	The changes a peptide carries relative to a reference protein it maps to; null when it matches its proteins as they are
proteoform	edits	sequence_edit	For a variant member, the changes that make it from the canonical entry, in canonical coordinates

Both columns are optional.

Substitutions and modifications

Some substitutions have the same mass as a modification: D→E and G→A weigh the same as methylation, and N→D the same as deamidation. A spectrum that fits both readings is a case of competing interpretations: a single-residue substitution can be listed as a candidate at its position, using Unimod's amino-acid substitution entries (e.g. Asp->Glu), next to the modification it competes with. The conventions for substitution candidates fix their coordinates, direction and assigned. Edits that span several residues cannot sit at one position and are recorded only as sequence edits.

Validation

Under qpxc validate (strict), each edit has start >= 1, a reference of end - start + 1 one-letter residues (empty with end = start - 1 for an insertion), an alternative of one-letter residues, and changes something (reference and alternative differ, so a no-op such as R→R fails). The edits of one sequence (a proteoform row, or one protein mapping of a pepmap row) are in order and do not overlap (conflicting_sequence_edits). Fields of the wrong type are reported as a type mismatch and the edit checks are skipped.