Regex Syntax

Motif parsing supports a controlled regex-like subset.

• Fixed residues from the selected alphabet:
  • protein (alphabet=ProteinAlphabet(), default): ARNDCQEGHILKMFPSTWYV
  • DNA (alphabet=DNAAlphabet()): ACGT
  • RNA (alphabet=RNAAlphabet()): ACGU
• Wildcards:
  • . is the canonical wildcard in normalized and output motifs.
  • x and X are accepted input aliases and mean the same thing as . in this package.
• Character classes:
  • [KR] includes listed residues.
  • [^P] is negation within the selected alphabet only.
  • repeated residues inside positive classes are ignored, so [KKAQ], [AQK], and [QQAAK] all normalize to the same residue set.
• Anchors:
  • ^ and $ indicate N- and C-terminus for protein motifs, or 5' and 3' ends for nucleic acid motifs.
• Repeat quantifiers:
  • {n}, {m,n} on residues, classes, anchors, and oracle-style grouped alternatives (see below).
• Grouping and alternation:
  • (...) for grouping and | for alternatives, for example A(K|Q)LI.
• Whitespace:
  • leading and trailing whitespace is trimmed;
  • the first internal whitespace ends motif parsing, matching the upstream oracle.

Edge Cases and Oracle Behavior

For a handful of parser edge cases, this package intentionally matches the observed black-box behavior of the upstream CompariMotif oracle rather than enforcing a stricter regex grammar.

• Internal whitespace truncates parsing at the first space, so A C behaves as A.
• Some permissive exact quantifiers are accepted and normalized the same way as the oracle, for example A{0} and A{-1} both behave as A.
• Grouped exact quantifiers are supported with the oracle's branch-splitting behavior rather than standard regex repetition semantics. For example, (A|C){2} expands to AA and CC, while (AC|GT){2} expands to ACC and GTT (equivalently normalized as (AC{2})|(GT{2})).
• The upstream black-box oracle has a grouped-alternation quirk where top-level wildcard-only branches such as (Q|.) and (Q|.){2} are dropped entirely, while top-level (Q|x) and (Q|X) retain the Q branch and only lose the wildcard-alias branch; embedded forms like A(Q|x)L still expand. This package intentionally does not reproduce that behavior. Instead, canonical normalization always treats x, X, ., and explicit full-alphabet classes as the same wildcard syntax. For example, (Q|[ARNDCQEGHILKMFPSTWYV]) normalizes as (Q)|(.).
• The upstream oracle also scores duplicate residues inside positive classes, but this package intentionally treats character classes as sets. For example, [AA] and [A] are equivalent here even though the oracle can distinguish them.
• Some malformed constructs are treated as non-retained motifs rather than hard parse failures, such as A{2,1} producing zero retained variants.
• Truly malformed syntax is still rejected when the oracle also rejects it, for example an unclosed character class like A[Q.