nx_supplement_screen.nx

nx_supplement_screen.nx -- SUPPLEMENT ARC / PEPTIDE ADULTERANT rung. Screens an observed LC-MS peak against a catalog of peptide actives sold in supplement and "research peptide" channels, ACROSS CHARGE STATES. WHY THIS EXISTS -- a measured hole, not a hypothetical one. nx_chem_adulterant_db screens small molecules well: it parses SMILES, computes a monoisotopic mass, and matches an observed peak against [M+H]+ within a ppm window. That design silently cannot see peptides: 1. A peptide is a SEQUENCE, not a SMILES the parser will take. 2. Electrospray puts peptides on a CHARGE LADDER. BPC-157 is 1418.70 Da, so [M+H]+ sits at m/z 1419.7 -- outside the m/z 100-1000 window a small-molecule method typically scans. Its real signals are [M+2H]2+ at 710.4 and [M+3H]3+ at 473.9. A screen that computes only [M+H]+ therefore reports CLEAN on a spiked sample. sup_mh_plus_in_range and sup_detectable_charges_in_range make that failure MECHANICAL rather than rhetorical: they compute, per compound, whether the single-charge assumption would have missed it. MASSES ARE DERIVED, NOT TABULATED. Every catalog mass comes from nx_peptide walking the actual sequence, plus an explicit terminal modification delta. Nothing here is a copied number that can drift out of agreement with the residue table. FAIL-CLOSED ON WHAT IT CANNOT DO. Peptides carrying non-proteinogenic residues (ipamorelin's Aib and 2-Nal, melanotan-II's Nle) are REFUSED by nx_peptide rather than approximated, so they screen as "unknown", never as "absent". Extending to them needs a non-standard residue table and is an honest open gap, listed below -- not quietly rounded away. All INTEGER. _q4 = x10^4 (mass/mz in Da), _q1 = x10 (ppm). Grounding (cited; researcher-groundable): fda_tainted_supplements_database (adulteration channel) wada_prohibited_list_s2_peptide_hormones (status flags) fenn_1989_electrospray_multiple_charging (the charge ladder) usp_dietary_supplement_label_claim_bands (dose acceptance) Open gaps (stated, not hidden):