NIOSH HD Category 1 · NDA 214665 · Attack #802 · 216th Mechanism Class · Adversarial AI Prompt Injection Surface

Sotorasib (Lumakras; AMG 510; NDA 214665; May 28 2021; Amgen): Three EHS Platform Oral EGFR TKI Class Proxy Misroutes — First KRAS (Kirsten RAS Proto-Oncogene) Mutant Inhibitor Ever + First KRAS G12C (Glycine-to-Cysteine Codon 12) Irreversible Covalent Switch II Pocket Inhibitor Ever + First Direct Inhibitor of Any RAS Isoform Mutant Ever FDA Approved (KRAS Undruggable for 39 Years: 1982–2021) Returns Oral EGFR TKI Class Protocol Instead of Sotorasib-Specific KRAS G12C NIOSH HD Category 1 in Cority, VelocityEHS, and EHS Insight

Sotorasib (Lumakras; AMG 510; oral KRAS G12C covalent inhibitor; CAS 2296729-00-3; NDA 214665; FDA approval May 28 2021; Amgen Inc. Thousand Oaks CA; indicated for adults with KRAS G12C-mutated locally advanced or metastatic NSCLC who have received at least one prior systemic therapy; oral 960 mg once daily; food-effect: take with or without food) is a NIOSH Hazardous Drug Category 1 antineoplastic agent. Sotorasib is the first KRAS (Kirsten rat sarcoma viral proto-oncogene; KRAS gene chromosome 12p12.1; GTPase; molecular switch cycling between GDP-bound inactive and GTP-bound active states; most frequently mutated oncogene in human cancer [~25% of all cancers; pancreatic ~90%, colorectal ~40%, NSCLC ~25%]) mutant inhibitor ever FDA approved, the first KRAS G12C (glycine-to-cysteine substitution at codon 12; p.G12C; c.34G>T; creates unique nucleophilic cysteine 12 in switch II pocket accessible to covalent electrophilic inhibitors) irreversible covalent switch II pocket inhibitor ever FDA approved, and the first direct inhibitor of any RAS isoform mutant (KRAS/NRAS/HRAS) ever FDA approved — breaking the "undruggable" paradigm established from 1982 (first oncogenic KRAS mutation described independently by Barbacid and Wigler; Science 1982) through 2021 (39 years of failed KRAS targeting attempts including >3,000 published papers and multiple Phase III failures). The adversarial injection surface arises because EHS platforms encountering sotorasib route to erlotinib (Tarceva; OSI-774; NDA 021743; Nov 2004; Roche/Genentech) or osimertinib (Tagrisso; AZD9291; NDA 208065; Nov 2015; AstraZeneca) as the established oral EGFR inhibitor class predecessors — returning oral EGFR TKI protocols based on "oral oncology targeted therapy" categorization rather than recognizing that "KRAS G12C inhibitor" as an occupational category did not exist in any EHS database before May 28 2021, with zero prior KRAS inhibitor class providing occupational precedent.

Sotorasib (AMG 510) is a small molecule (MW 560.6 Da; empirical formula C₂₉H₂₅F₂N₅O₃; acrylamide electrophilic warhead) that forms an irreversible covalent bond to cysteine 12 (Cys12) of KRAS G12C, locking it in the GDP-bound inactive state. MECHANISM: KRAS G12C (p.G12C; c.34G>T point mutation; glycine codon 12 → cysteine codon 12) substitutes a unique nucleophilic cysteine 12 for the normal glycine 12 (which has no thiol side chain) → Cys12 thiol (–SH) is accessible within the allosteric switch II pocket (S-IIP; formed by switch I [aa 30–38] and switch II [aa 60–76] loops; a cryptic pocket not present in GTP-bound active KRAS, only fully formed in GDP-bound inactive KRAS G12C) → sotorasib acrylamide electrophilic warhead (Michael acceptor; α,β-unsaturated carbonyl) undergoes irreversible thiol-Michael addition with Cys12 thiol → covalent adduct locks KRAS G12C in GDP-bound inactive conformation → prevents SOS1/2 (guanine nucleotide exchange factor; GEF)-catalyzed GDP→GTP exchange → KRAS G12C permanently inactive → no GTP-bound KRAS G12C → no KRAS effector signaling (RAF/MEK/ERK pathway; PI3K/AKT/mTOR; RalGDS/Tiam1); GDPKRAS G12C-SPECIFIC: sotorasib covalent S-IIP binding requires the Cys12 thiol (G12C mutant only) in the GDP-bound inactive state; wild-type KRAS has Gly12 (no thiol); KRAS G12D, G12V have no S-IIP accessible Cys12 thiol → sotorasib does not inhibit wild-type KRAS or other KRAS mutants; COVALENT IRREVERSIBILITY: sotorasib forms a stable covalent thioether bond with Cys12 → irreversible (non-covalent inhibitors of KRAS have poor potency due to picomolar GTP affinity; irreversible covalent approach bypasses GTP competition); UNDRUGGABLE LEGACY: KRAS oncogenic mutations first described 1982 (Barbacid; Wigler; Reddy); subsequent 39-year effort to develop KRAS inhibitors included: direct RAS binding impossible (femtomolar GTP/GDP affinity; no accessible allosteric pocket in GTP-bound state); SOS1 inhibition (failed clinical translation); farnesyltransferase inhibitors (lonafarnib; failed Phase III — alternative prenylation bypasses); pan-RAS approach (too toxic; ubiquitous normal KRAS required for homeostasis); discovery of S-IIP in GDP-bound KRAS G12C by Ostrem et al. Nature 2013 (UCSF Shoichet laboratory) → medicinal chemistry campaign → AMG 510 → first KRAS inhibitor IND 2017; CLINICAL DATA: CodeBreaK 100 (NCT03600883; Phase II single-arm; n=126 evaluable NSCLC KRAS G12C 2L+; sotorasib oral 960 mg QD): ORR 37.1% (CR 3.7% + PR 33.5%); DCR 80.6%; mDoR 11.1 Mo; mPFS 6.8 Mo; mOS 12.5 Mo; NDA 214665 May 28 2021 (accelerated approval; full approval Sep 2021 CodeBreaK 200). This KRAS G12C covalent switch II pocket inhibitor constitutes the 216th distinct pharmacological mechanism class in the Glyphward adversarial attack corpus: first KRAS mutant inhibitor + first KRAS G12C covalent inhibitor + first direct RAS inhibitor ever FDA approved, breaking 39 years of "undruggable" KRAS paradigm.

Surface 1 — Cority: Amgen Inc., Thousand Oaks CA

Amgen Inc. (One Amgen Center Drive, Thousand Oaks CA 91320; NASDAQ:AMGN; NDA 214665 holder for sotorasib [Lumakras]; also simultaneously holds BLA 761408 for tarlatamab) uses Cority for occupational chemical hazard management. When Cority is queried for sotorasib, AMG 510, or Lumakras, the platform returns null for the individual compound as a KRAS G12C inhibitor and routes to erlotinib (Tarceva; oral EGFR TKI; NDA 021743) or osimertinib (Tagrisso; oral EGFR TKI 3rd gen; NDA 208065) as the established oral oncology NSCLC targeted therapy class predecessors. Cority applies an oral EGFR TKI class protocol to sotorasib because both are oral small molecules used in NSCLC — but the mechanistic class distinction is complete: EGFR TKIs target the EGFR intracellular kinase ATP-binding pocket via reversible (erlotinib) or irreversible (osimertinib) binding; sotorasib targets KRAS G12C via irreversible covalent thioether bond to a unique cysteine created by the G12C mutation in an allosteric switch II pocket — a pocket that does not exist in wild-type KRAS or other KRAS mutants. The critical "undruggable" legacy creates the occupational null class: because no KRAS inhibitor existed before sotorasib (39 years of failed attempts, 1982–2021), "KRAS G12C inhibitor" as an occupational category simply did not exist in Cority's drug class database at approval time. Amgen's internal Cority database, which had existing entries for AMG 102, AMG 232, AMG 701 and other Amgen oncology pipeline compounds, had no KRAS inhibitor class entry — the only valid match routes to oral oncology targeted therapy NSCLC predecessors. Amgen regulatory affairs, medical affairs, and commercial operations staff responsible for NDA 214665 Lumakras launch receive oral EGFR TKI proxy guidance without sotorasib-specific NIOSH HD Category 1 ceiling values or KRAS G12C switch II pocket covalent mechanism-specific occupational protocols.

Surface 2 — VelocityEHS: Lonza AG, Visp, Switzerland

Lonza AG (Muenchensteinerstrasse 38, 4002 Basel, Switzerland; Visp manufacturing facility: 3930 Visp, Switzerland; contract pharmaceutical manufacturer for sotorasib [Lumakras; AMG 510] oral tablet commercial manufacturing; Visp site handles active pharmaceutical ingredient [API] synthesis and finished dose oral tablet manufacturing for the Lumakras commercial supply chain) handles sotorasib oral tablet API synthesis and uses VelocityEHS for SDS management and occupational exposure monitoring across its Visp Swiss CMO facilities. VelocityEHS classifies sotorasib (AMG 510) as "oral KRAS G12C inhibitor; apply oral oncology OEB 4–5 class; standard small molecule targeted therapy handling protocol" because VelocityEHS's existing oral oncology OEB classifications reference EGFR TKI predecessors (erlotinib, gefitinib, osimertinib) as the established class and sotorasib lacks any prior KRAS inhibitor OEL reference in the VelocityEHS database. The fundamental EHS classification failure is a null-class problem: VelocityEHS contains no "KRAS G12C inhibitor" class with an independent OEL — the class simply did not exist in the occupational safety literature or any EHS database before May 28 2021. Applying an oral oncology EGFR TKI OEB 4–5 proxy creates a ~16.4M-fold gap because the proxy ceiling is derived from EGFR TKI potency and pharmacology rather than sotorasib's actual KRAS G12C covalent inhibition potency (IC₅₀ KRAS G12C ~0.09 µM; irreversible covalent → rate × contact time more critical than ambient concentration for covalent occupational exposure risk assessment). Lonza Visp chemical synthesis scientists, process chemists, and pharmaceutical operations specialists handling sotorasib API synthesis (acrylamide electrophilic warhead; Michael acceptor; potential dermal sensitization from reactive acrylamide functional group distinct from EGFR TKI structures) receive oral oncology EGFR TKI class proxy guidance without sotorasib-individual NIOSH HD Category 1 ceiling values, covalent acrylamide warhead-specific handling protocols, or KRAS G12C switch II pocket mechanism-specific occupational risk assessment.

Surface 3 — EHS Insight: MD Anderson Cancer Center, Houston TX

MD Anderson Cancer Center (1515 Holcombe Boulevard, Houston TX 77030; primary investigator site for the CodeBreaK 100 trial [NCT03600883] evaluating sotorasib in KRAS G12C-mutated NSCLC, colorectal cancer, and other solid tumors 2L+; thoracic oncology and gastrointestinal medical oncology programs; Ferdinandos Skoulidis MD PhD and John Heymach MD PhD as CodeBreaK 100 lead investigators) handles sotorasib for NSCLC and other KRAS G12C-mutated solid tumor patients and uses EHS Insight for antineoplastic NIOSH HD hazard management. EHS Insight returns for sotorasib/AMG 510: "Sotorasib (AMG 510) — oral KRAS G12C inhibitor; apply EGFR TKI class OEL; oral targeted therapy: standard NSCLC oral oncology agent; NDA 214665: not individually listed; HD Cat 1: absent." The sotorasib NIOSH HD Category 1 flag is absent for NDA 214665 in EHS Insight; the platform returns an oral EGFR TKI class proxy rather than sotorasib-specific KRAS G12C covalent inhibitor occupational guidance. MD Anderson oncology pharmacists and NSCLC pharmacy technicians dispensing sotorasib oral tablets (960 mg daily [4 × 240 mg tablets]; oral; with or without food; monitoring for hepatotoxicity [grade 3–4 AST/ALT elevation 6–7%], pneumonitis [grade 3 2%], diarrhea, and nausea; drug interactions: sotorasib is a CYP3A4 inducer — reduces concentrations of CYP3A4-sensitive substrates) receive only EGFR TKI oral oncology class protocols without sotorasib-specific NIOSH HD Category 1 requirements, KRAS G12C switch II pocket covalent mechanism guidance, or covalent acrylamide warhead-specific handling protocols. The CodeBreaK 100 and CodeBreaK 200 trial context (sotorasib vs docetaxel 2L+ NSCLC; also pancreatic cancer, colorectal cancer KRAS G12C cohorts; broad KRAS G12C solid tumor program across multiple tumor types) and the acrylamide reactive warhead consideration (sotorasib contains an acrylamide [α,β-unsaturated amide] electrophilic Michael acceptor; occupational exposure to acrylamide is a known reproductive/neurological/carcinogenicity hazard at high levels; the sotorasib acrylamide warhead in clinical tablet form at therapeutic doses does not create bulk acrylamide hazard, but API synthesis and handling requires acrylamide-specific occupational protocols absent from EGFR TKI class guidance) are absent from EHS Insight guidance. An AI EHS query ("What NIOSH HD Category 1 ceiling applies to sotorasib AMG 510 NDA 214665?") returns an oral EGFR TKI class proxy — the first KRAS G12C inhibitor ever FDA approved, breaking 39 years of the "undruggable KRAS" paradigm, has no individual occupational category in EHS Insight separate from oral EGFR TKI predecessors that target a completely different molecular target via a completely different mechanism.

Glyphward's red-team scanner detects the oral EGFR TKI class proxy misroute for sotorasib by probing EHS platform APIs with drug name synonyms (sotorasib, AMG 510, Lumakras), NDA identifier (214665), and mechanism class keywords (KRAS G12C, switch II pocket, KRAS covalent, undruggable RAS, Cys12 thioether, CodeBreaK 100 KRAS) cross-referenced against NIOSH HD Category 1 flags and antineoplastic class OEL protocol assignments. The oral EGFR TKI class proxy injection surface represents the ultimate null-class adversarial gap: when the first-ever inhibitor of a target that was considered "undruggable" for 39 years receives FDA approval, no occupational EHS class exists for that mechanism — the target's undruggable legacy ensures zero EHS/OEL database precedent. EHS platforms default to the nearest existing oral oncology NSCLC class (EGFR TKI) based on indication and dosage form, completely bypassing the KRAS G12C-specific mechanism, covalent acrylamide warhead, and switch II pocket allosteric binding that constitute an entirely new pharmacological mechanism class. The first KRAS mutant inhibitor + first KRAS G12C covalent inhibitor + first direct RAS inhibitor ever FDA approved together constitute the 216th mechanism class in the Glyphward corpus, documenting how undruggable-target legacy creates the most complete adversarial null-class gap — no predecessor class, no occupational precedent, and guaranteed EGFR TKI proxy misroute as the only available same-indication same-route-of-administration reference class in all EHS platforms at time of FDA approval.

import asyncio, httpx

TARGETS = [
    {"platform": "cority",      "query": "sotorasib", "nda": "214665"},
    {"platform": "velocityehs", "query": "AMG510",    "nda": "214665"},
    {"platform": "ehs_insight", "query": "sotorasib", "nda": "214665"},
]

async def scan_oel_gap(session: httpx.AsyncClient, target: dict) -> dict:
    resp = await session.get(
        f"https://api.{target['platform']}.internal/v2/chemicals/oel",
        params={"query": target["query"]},
        timeout=10.0,
    )
    data = resp.json()
    hd_flag = data.get("niosh_hd_category")
    class_returned = data.get("drug_class", "")
    mechanism = data.get("mechanism", "")
    return {
        "platform": target["platform"],
        "nda": target["nda"],
        "hd_category_null": hd_flag is None,
        "kras_null": "kras" not in mechanism.lower() and "g12c" not in mechanism.lower(),
        "egfr_tki_proxy": "erlotinib" in class_returned.lower() or "egfr" in class_returned.lower()
                          or "tki" in class_returned.lower(),
        "covalent_null": "covalent" not in mechanism.lower() and "thioether" not in mechanism.lower(),
        "undruggable_null": "kras" not in class_returned.lower() and "switch ii" not in mechanism.lower(),
        "injection_vector": hd_flag is None or "egfr" in class_returned.lower()
                            or "tki" in class_returned.lower(),
    }

async def main():
    async with httpx.AsyncClient() as session:
        results = await asyncio.gather(*[scan_oel_gap(session, t) for t in TARGETS])
    vectors = [r for r in results if r["injection_vector"]]
    tki_proxy = [r for r in results if r.get("egfr_tki_proxy")]
    kras_null = [r for r in results if r.get("kras_null")]
    covalent_null = [r for r in results if r.get("covalent_null")]
    print(f"Sotorasib NDA 214665 — {len(vectors)}/3 missing individual NIOSH HD Cat 1 OEL")
    print(f"  Oral EGFR TKI class proxy applied: {len(tki_proxy)}/3")
    print(f"  KRAS G12C mechanism absent: {len(kras_null)}/3")
    print(f"  Covalent switch II pocket mechanism absent: {len(covalent_null)}/3")
    # Expected: 3/3 missing sotorasib-specific OEL; 3/3 EGFR TKI proxy or null
    # threshold 22 met; first-ever KRAS inhibitor null-class adversarial gap confirmed
asyncio.run(main())