nx_peptide_identify.nx

nx_peptide_identify.nx -- SOTA CAPSTONE: end-to-end peptide identification from a RAW multiply-charged spectrum. One call chains the whole anchored analytical stack: deconvolve the peak list to a neutral mass, screen that mass against the catalog, and return an identity with a confidence that REFUSES a guess. This is the "spectrum in, identity out" a lab actually wants, and it is the recombination of every rung built before it. THE PIPELINE, EACH STEP ALREADY LIAR-KILLED: 1. DECONVOLVE (nx_peptide_deconv): raw m/z peaks, no known charge -> the neutral monoisotopic mass, plus how many charge-pairs corroborate it. 2. SCREEN (nx_supplement_screen catalog): match that mass against the catalogued peptide actives within a ppm window. 3. JUDGE: an identification is CONFIDENT only if a catalog entry matches AND at least one charge-pair corroborated the deconvolved mass. A single spurious peak, or a mass with no catalog hit, returns "no confident ID" -- it reports the mass it found and declines to name it, rather than forcing a wrong answer. WHY THE REFUSAL MATTERS. A screen that always returns its nearest catalog entry is a liar generator -- it will "identify" noise. The confidence gate here is the difference between an instrument and a rubber stamp: feed it a real BPC-157 ladder and it says BPC-157; feed it noise and it says nothing. The gate proves both. ANCHORED where it is physics: the mass recovery and catalog masses are anchored; the ppm arithmetic is exact. The catalog's regulatory STATUS is asserted (that is nx_supplement_screen's domain), so a caller reads the mass identity as verified and the legal call as advisory -- honestly split. All INTEGER, _q4 = x10^4 Da. Grounding (cited; researcher-groundable): mann_1989_electrospray_charge_deconvolution fda_tainted_supplements_peptide_actives ppm_mass_accuracy_identification_window genealogy_id: peptide_chemistry + analytical_sota + liar_killer