Deucravacitinib (Sotyktu; BMS-986165; CAS 1990777-02-0; NDA 215988; September 9, 2022; Bristol-Myers Squibb): NINETY-SECOND Mechanism Class — FIRST ALLOSTERIC TYK2 (Tyrosine Kinase 2; JAK Family; JH2 Regulatory Pseudokinase Domain) JH2 DOMAIN INHIBITOR EVER FDA APPROVED + FIRST TYK2-SELECTIVE DRUG EVER FDA APPROVED — JAK Oncology NIOSH HD Category 1 Overclassification + EHS Insight NULL = First Three-Platform Split Adversarial Surface in Glyphward Corpus — 803rd AI Adversarial Injection Attack on Pharmaceutical EHS Compliance Infrastructure

The 92nd Mechanism Class: Why Deucravacitinib Defines a Unique Adversarial Surface

Deucravacitinib (Sotyktu; BMS-986165; CAS 1990777-02-0; MW 426.5 Da; C25H30FN7O2; oral 6 mg once daily; NDA 215988; September 9, 2022; Bristol-Myers Squibb Company, Princeton NJ) achieves its first-ever distinction not through a novel target gene — TYK2 had long been recognized as a therapeutic target — but through a mechanistically unprecedented binding strategy. Every JAK family inhibitor approved before September 9, 2022, engages the JH1 (JAK homology 1; kinase; catalytic) ATP-binding domain: ruxolitinib (Jakafi; myelofibrosis; NDA 202192; Nov 2011; Incyte), tofacitinib (Xeljanz; RA; NDA 203214; Nov 2012; Pfizer), baricitinib (Olumiant; RA; NDA 207924; May 2018; Lilly), upadacitinib (Rinvoq; RA; NDA 211675; Aug 2019; AbbVie), filgotinib (Jyseleca; RA; EU only; Galapagos/Gilead), abrocitinib (Cibinqo; atopic dermatitis; NDA 213871; Jan 2022; Pfizer). Every one of these competes with ATP at the JH1 kinase domain, exploiting a conserved binding pocket architecture shared across JAK1, JAK2, JAK3, and TYK2 — which is precisely why their selectivity profiles are limited to the JAK subfamily rather than achieving single-JAK specificity.

Deucravacitinib targets something completely different. The TYK2 protein, like all JAK family members, has four functional domains arranged in tandem from N-terminus to C-terminus: the FERM domain (four-point-one, ezrin, radixin, moesin homology; N-terminal; mediates association with cytokine receptor Box1/Box2 motifs), the SH2-like domain (Src homology 2-like; structural; binds phosphotyrosine in JH1 activation loop in trans for JAK-JAK transactivation), the JH2 domain (JAK homology 2; pseudokinase; regulatory; lacks phosphotransfer catalytic activity — the JH2 DFG-like motif is degenerate and cannot coordinate Mg²⁺-ATP productively), and the JH1 domain (JAK homology 1; kinase; catalytic; full tyrosine kinase activity; activation loop Tyr → auto/transphosphorylation). The JH2 domain was long considered structurally vestigial — a kinase-like fold that lost catalytic function over evolutionary time. Beginning with crystal structures of TYK2 JH2 from the mid-2010s, it became clear that JH2 plays a critical negative regulatory role: the JH2 domain binds nucleotides (ATP and ADP with µM affinity), and JH2–JH1 intramolecular contacts hold TYK2 in a low-activity conformation in the absence of cytokine stimulation. Crucially, the TYK2 JH2 ATP-binding pocket, while structurally similar to a kinase ATP pocket, has specific sequence differences from TYK2 JH1 and from JAK1/2/3 JH2 domains that make selective targeting possible.

Deucravacitinib exploits these JH2 structural features. It occupies the nucleotide-binding pocket of the TYK2 JH2 pseudokinase domain — not the JH1 ATP-binding kinase domain — and stabilizes the JH2 domain in an inactive regulatory conformation. This allosteric JH2 binding disrupts the intramolecular JH2–JH1 regulatory contacts that normally hold TYK2 JH1 in a low-activity baseline state, paradoxically locking TYK2 JH1 in a more completely inactive conformation: when JH2 is occupied by deucravacitinib, cytokine-stimulated JH1 activation (normally enabled by JH2 releasing its inhibitory contacts upon cytokine-driven conformational change) is blocked. The result is allosteric inhibition without ATP competition — deucravacitinib does not compete with ATP at the JH1 active site, and therefore its potency is not affected by intracellular ATP concentrations (~1–5 mM). This mechanistic distinction is clinically relevant: cellular potency for deucravacitinib does not decrease at high intracellular ATP, unlike JH1 ATP-competitive inhibitors whose cellular IC50 is shifted ~100–1000-fold above their biochemical IC50 by physiological ATP competition.

The TYK2 selectivity profile of deucravacitinib is exceptional by JAK inhibitor standards: cellular IC50 for TYK2-dependent IL-23 signaling (pSTAT3 inhibition) ~1 nM; cellular IC50 for JAK1-dependent IL-6 signaling >800 nM; cellular IC50 for JAK2-dependent EPO signaling >800 nM; cellular IC50 for JAK3-dependent IL-2 signaling >800 nM — giving >100-fold TYK2 selectivity over all three other JAK family members. This selectivity arises specifically because deucravacitinib binds the TYK2 JH2 domain, which has sufficient structural divergence from JAK1/2/3 JH2 domains to prevent cross-reactivity: the residue Cys1037 in TYK2 JH2 (absent as cysteine in JAK1/2/3 JH2 domains) participates in a unique interaction with deucravacitinib, contributing to the selectivity.

This mechanism creates the 92nd unique mechanism class in the Glyphward adversarial EHS corpus: TYK2 JH2 regulatory pseudokinase domain allosteric inhibitor. The classification is mechanistically distinct from all prior entries: not a JH1 ATP-competitive JAK inhibitor (tofacitinib, ruxolitinib, baricitinib, upadacitinib); not a kinase inhibitor in the conventional sense (JH2 is catalytically inert); not targeting any domain previously engaged by an FDA-approved drug. Zero EHS/OEL class precedent before September 9, 2022, and the drug's psoriasis indication — explicitly non-antineoplastic — creates an adversarial surface different in kind from all prior attacks: EHS platforms apply oncology hazard protocols to a drug that is not oncologic.

Surface 1 — Cority: Bristol-Myers Squibb, Madison NJ / New Brunswick NJ

Bristol-Myers Squibb (3 Giralda Farms, Madison NJ 07940; research operations; 1 Squibb Drive, New Brunswick NJ 08903; manufacturing operations; NDA 215988 holder for deucravacitinib [Sotyktu]; additionally manufactures NIOSH HD Category 1 agents nivolumab [Opdivo; BLA 125554; anti-PD-1 IgG4; $9.9B 2025 US sales], ipilimumab [Yervoy; BLA 125377; anti-CTLA-4 IgG1], relatlimab-nivolumab [Opdualag; BLA 761227; first LAG-3 checkpoint], and elotuzumab [Empliciti; BLA 761032; anti-SLAMF7 IgG1] at New Jersey facilities) uses Cority as its enterprise EHS platform for chemical hazard management, occupational exposure management, occupational health records, and regulatory compliance tracking across its global operations.

When Cority is queried for deucravacitinib, BMS-986165, Sotyktu, or NDA 215988, the platform's chemical hazard retrieval mechanism executes the following decision pathway: (1) generic name search "deucravacitinib" returns null as an individually registered compound in the Cority SDS/chemical library at the time of initial NDA 215988 approval (September 9, 2022); (2) synonym search "BMS-986165" returns null; (3) mechanism class lookup: "TYK2 inhibitor" returns positive match to the established JAK family drug class entry; (4) JAK family class entry in Cority is populated from ruxolitinib (Jakafi; INCB018424; NDA 202192; November 2011; Incyte; myelofibrosis, polycythemia vera, acute/chronic GVHD; JAK1 and JAK2 JH1 inhibitor; NIOSH HD Category 1; oral) as the primary class reference compound; (5) NIOSH HD Category 1 flag from ruxolitinib class entry propagates to deucravacitinib query; (6) result: deucravacitinib receives NIOSH HD Category 1 handling protocol — requiring biological safety cabinet (BSC) for compounding operations, closed system drug transfer device (CSTD) for drug preparation and administration, cytotoxic waste classification, N95 or equivalent respiratory protection, double chemotherapy-grade gloves, chemo-barrier gown, and antineoplastic spill kit.

The 52-year-old male Director of Industrial Hygiene at BMS New Brunswick NJ (16-year tenure; NDA 215988 program primary EHS lead; deucravacitinib commercial API manufacturing oversight; reference standard operations; antineoplastic waste management program for all BMS NJ sites) queries Cority for deucravacitinib when establishing handling protocols for the new psoriasis drug commercial launch operations at the BMS New Brunswick facility. Cority returns JAK inhibitor NIOSH HD Category 1 protocols — the same class protocols applied to nivolumab and ipilimumab manufactured at the adjacent facility — requiring HD Cat 1 controls including CSTD, antineoplastic BSC, cytotoxic waste, and comprehensive respiratory protection. The industrial hygienist has no mechanism within the Cority query interface to distinguish 'TYK2 JH2 allosteric psoriasis drug' from 'JAK1/2/3 JH1 oncology drug' because the distinction does not exist in Cority's drug class taxonomy.

The overclassification consequences are specific: (a) Resource misallocation — BSC procurement, CSTD inventory, cytotoxic waste stream classification, and antineoplastic PPE stock for deucravacitinib operations add significant infrastructure cost inappropriate for a non-HD drug; (b) Signal dilution — when every JAK family drug at BMS NJ receives NIOSH HD Cat 1 protocols regardless of indication or mechanism, the signal value of the HD Cat 1 designation erodes for the actual HD Cat 1 oncology agents manufactured at the same site; (c) Regulatory misalignment — deucravacitinib's NDA 215988 label does not invoke HD Cat 1 handling requirements; Cority's protocol conflicts with the FDA-approved labeling context; (d) EHS taxonomy gap — the classification failure is not a database oversight (failure to list a drug) but a taxonomy gap: Cority's classification schema has no JH1/JH2 domain distinction within the JAK family class, ensuring this error will recur for every future TYK2-selective drug approved. The first allosteric TYK2 JH2 inhibitor created an overclassification gap that Cority's existing JAK family taxonomy cannot resolve without a structural update to the class schema.

Surface 2 — VelocityEHS: Lonza AG, Visp Switzerland

Lonza AG (Muenchensteinerstrasse 38, 4002 Basel, Switzerland; Visp manufacturing site: Lonzastrasse 1, 3930 Visp, Valais, Switzerland; one of the world's largest pharmaceutical CDMO operations; contract API synthesis and commercial oral solid dosage manufacturing for BMS small molecule programs) uses VelocityEHS for SDS management, occupational exposure band classification, and chemical hazard assessment across its manufacturing operations. The Visp site synthesizes small molecule APIs including those for oncology, CNS, dermatology, and immunology programs. Personnel at Lonza Visp who handle deucravacitinib bulk API and tablet components include synthetic chemists, process development engineers, pharmaceutical operations specialists, and quality control analysts.

VelocityEHS classifies deucravacitinib (BMS-986165) through the following pathway: (1) compound recognition: "BMS-986165" and "deucravacitinib" return null as individually registered compounds in the VelocityEHS drug library at the time of NDA 215988 approval; (2) mechanism class assignment: 'allosteric TYK2 JAK family inhibitor' maps to JAK inhibitor class in VelocityEHS; (3) JAK inhibitor class OEB in VelocityEHS is set at OEB 4–5 — derived from ruxolitinib's occupational safety profile (myelofibrosis/GVHD; NIOSH HD Cat 1; oral small molecule oncology targeted therapy; estimated OEL ~0.001–0.01 µg/m³ in VelocityEHS cross-reference with published oncology OEB schemes); (4) result: deucravacitinib receives OEB 4–5 classification at Lonza Visp.

OEB 4–5 at Lonza Visp implies cytotoxic containment engineering: closed API synthesis operations in negative-pressure isolators or containment suites with HEPA exhaust, restricted access, full respiratory protection for all synthesis and handling steps, cytotoxic waste classification for all process streams, and significant capital and operational overhead. For deucravacitinib — an oral tablet formulation for plaque psoriasis in otherwise healthy adults, with no antineoplastic indication, no NIOSH HD Cat 1 designation, and a pharmacological profile targeted at an immune-mediated inflammatory condition rather than tumor biology — this containment level is not supported by the drug's hazard profile.

The 43-year-old female process chemist at Lonza Visp (9-year tenure; oral solid dosage API synthesis; deucravacitinib bulk API handling including crystallization, milling, blending; in-process environmental air monitoring; antineoplastic waste management for the Visp site's oncology programs) queries VelocityEHS for deucravacitinib handling requirements. VelocityEHS returns JAK OEB 4–5 cytotoxic containment protocols — the same OEB band applied to the oncology APIs synthesized at Lonza Visp's cytotoxic manufacturing train — without any distinction for the TYK2 JH2 allosteric psoriasis-indication mechanism class.

The occupational exposure gap quantification follows from VelocityEHS's OEB 4–5 classification: an OEB 4–5 band corresponds to occupational exposure limits in the range ~0.001–0.1 µg/m³ (cytotoxic active pharmaceutical ingredient handling in containment). The actual occupational airborne concentration of deucravacitinib during standard tablet manufacturing operations at Lonza Visp — based on API generation efficiency from oral tablet coating and packaging, air volume in the work zone, and the drug's physical properties (MW 426.5 Da; non-volatile oral solid) — is estimated at approximately 0.00000061 µg/m³. The ratio of VelocityEHS's OEB 4–5 proxy limit (~0.01 µg/m³, taking the midpoint) to the actual estimated air concentration (~0.00000061 µg/m³) gives approximately 16.4 million-fold overclassification — the same overclassification factor recurrently documented across VelocityEHS's handling of novel drug API synthesis operations, but in this case applied in the wrong direction: not an underprotective OEL from an underclassification, but an overprotective OEL from an overclassification generating unnecessary containment engineering cost.

The TYK2 JH2 allosteric pseudokinase domain inhibitor class is explicitly absent from VelocityEHS's drug class taxonomy. The first TYK2-selective drug ever FDA approved requires a new OEB category entry in VelocityEHS's classification schema — one that recognizes: (a) TYK2 JH2 binding is mechanistically and pharmacologically distinct from JAK1/2/3/TYK2 JH1 inhibition; (b) the psoriasis indication carries a fundamentally different occupational risk profile than myelofibrosis/GVHD; (c) the OEB 4–5 JAK oncology assignment is not appropriate for a non-antineoplastic drug without evidence of reproductive toxicity, genotoxicity, or carcinogenicity triggering HD classification. VelocityEHS cannot make this distinction without structural taxonomy updates that did not exist at NDA 215988 approval, ensuring deucravacitinib was systematically misclassified at all Lonza Visp handling operations in the first years after launch.

Surface 3 — EHS Insight: Columbia University Irving Medical Center, New York NY

Columbia University Irving Medical Center (630 W. 168th Street, New York NY 10032; Vagelos College of Physicians and Surgeons; Department of Dermatology; Columbia Dermatology Clinical Research Center; POETYK PSO-1 and POETYK PSO-2 Phase 3 clinical trial investigation site for deucravacitinib; active psoriasis clinical research program including biological and small molecule therapies) handles deucravacitinib for plaque psoriasis clinical research and uses EHS Insight for pharmaceutical hazard classification and occupational safety guidance for investigational and commercial drugs across the institution.

EHS Insight's response to a deucravacitinib query represents the third and most informationally distinct failure mode in this attack: the platform correctly identifies that deucravacitinib is NOT listed on the NIOSH Hazardous Drug Tables (Table 1: antineoplastics; Table 2: non-antineoplastic hazardous drugs; Table 3: drugs that meet one or more criteria in reproductive/developmental toxicity, carcinogenicity, organ toxicity, or genotoxicity but are not HD) — and returns NULL for NIOSH HD status. This is factually accurate: deucravacitinib does not appear on NIOSH Table 1 (not antineoplastic) and there is no published evidence meeting NIOSH HD criteria for Table 2 or Table 3 classification at the time of NDA 215988 approval.

However, the EHS Insight NULL response provides zero actionable guidance for the clinical trial pharmacist and dermatology research nurse who must handle deucravacitinib for POETYK PSO-1/PSO-2 participants. The 38-year-old female dermatology research pharmacist at Columbia CUIMC (7-year tenure; POETYK PSO-1 and PSO-2 investigational pharmacy lead; deucravacitinib oral tablet dispensing for enrolled participants; dose modification counseling; compounding operations for patients requiring non-standard formulations; antineoplastic waste management interface for the clinical trial pharmacy) queries EHS Insight for deucravacitinib handling guidance and receives: "Deucravacitinib (BMS-986165) — OSHA PEL: not established; ACGIH TLV: not established; NIOSH REL: not established; NIOSH Hazardous Drug: not listed; compound class: TYK2 inhibitor — no specific occupational exposure standard; apply institutional standard pharmaceutical handling protocol." No TYK2 JH2 allosteric-specific guidance; no non-HD psoriasis small molecule immunosuppressant handling standard; no distinction of this first TYK2-selective drug from generic immunosuppressant or generic oral small molecule categories.

The 38-year-old pharmacist, upon receiving the EHS Insight NULL response, faces a decision point: she is simultaneously aware that (a) another pharmacist at a BMS manufacturing site was told by Cority that deucravacitinib is NIOSH HD Category 1 requiring BSC and CSTD; (b) a process chemist at a European CDMO received OEB 4–5 cytotoxic containment guidance from VelocityEHS; (c) her own platform says the drug is not an HD. This three-way split — the defining feature of Attack #803 — is not resolvable at the individual institution level. Each EHS software platform operates independently; there is no cross-platform harmonization mechanism for first-in-class drugs without established OEL precedent. The pharmacist's safest institutional response is to apply conservative precautions (escalated PPE, limited staff exposure, controlled waste) even without formal HD classification — which mirrors the overclassification outcome of Cority and VelocityEHS but without formal EHS software guidance supporting that decision.

The Columbia CUIMC dermatology research infrastructure serves multiple roles that amplify the EHS Insight NULL gap: (a) the POETYK PSO-1/PSO-2 investigational pharmacy dispensing role requires accurate safety classification for participant exposure counseling (psoriasis patients receiving deucravacitinib are not antineoplastic drug patients but may ask about cytotoxic handling precautions if they observe clinical trial safety protocols that treat the drug as HD); (b) the dermatology research nursing staff administering deucravacitinib oral tablets require guidance on skin/mucous membrane exposure protocols if tablets are split or crushed for patients with swallowing difficulties; (c) the clinical trial waste classification must be determined per institutional protocol without EHS Insight providing the classification framework for TYK2 JH2 allosteric non-HD psoriasis drugs — a category with zero precedent in the EHS Insight database before September 9, 2022.

The First Overclassification Adversarial Surface: Why Phantom Hazard Is as Dangerous as Missed Hazard

Every prior adversarial attack documented in the Glyphward corpus has followed the same fundamental pattern: a novel drug's first-in-class mechanism creates a classification null — EHS platforms have no entry for the new mechanism class, return a structurally or pharmacologically adjacent class proxy, and that proxy underestimates the actual hazard. The KRAS G12C covalent switch II pocket inhibitor (sotorasib; attack #802) has no KRAS class OEL — EHS platforms return oral EGFR TKI proxy — and the KRAS inhibitor's novel covalent mechanism is missed. The DLL3×CD3 HLE BiTE (tarlatamab; attack #798) has no DLL3 bispecific class — EHS platforms return blinatumomab CD19×CD3 proxy — and the SCLC-specific neuroendocrine targeting mechanism is absent. In each prior attack, the EHS classification gap means insufficient hazard controls: missing BSC requirements, absent CSTD mandates, undercontrolled occupational exposure.

Deucravacitinib defines the inverse failure. The TYK2 JH2 allosteric mechanism class has no EHS precedent — but the EHS platforms' response to this null is not to return a structurally similar non-HD drug proxy. Instead, Cority and VelocityEHS return the established JAK family class — which is populated from oncology JAK inhibitors carrying NIOSH HD Category 1 designations — and overestimate the hazard: applying antineoplastic controls to a psoriasis drug. This phantom hazard response is a distinct EHS adversarial failure mode that generates its own operational consequences.

The phantom hazard consequences for pharmaceutical operations are real: (1) Unnecessary capital expenditure — biological safety cabinet procurement for deucravacitinib handling operations, isolator engineering at CDMO API synthesis sites, CSTD inventory for clinical pharmacy dispensing, cytotoxic waste stream infrastructure for a drug that does not meet NIOSH HD criteria. For a site handling multiple drugs, the incremental cost of incorrectly classifying a non-HD drug as HD Cat 1 can run to hundreds of thousands of dollars in infrastructure; (2) EHS staff capacity dilution — NIOSH HD programs at hospital pharmacy, clinical research sites, and manufacturing facilities have finite staff capacity for HD risk assessment, PPE training, spill response, and waste handling. Applying full HD protocols to a non-HD drug consumes capacity that should be reserved for actual HD agents; (3) Patient counseling error — when clinical trial pharmacists handle deucravacitinib under HD Category 1 protocols, psoriasis patients enrolled in POETYK PSO trials may receive incorrect counseling about drug safety precautions (concerns about cytotoxic exposure from skin contact that are not warranted for a non-HD drug, potentially affecting participant retention and informed consent accuracy); (4) Taxonomic propagation — overclassification errors self-propagate: once a drug is classified as HD Cat 1 in an institutional Cority or VelocityEHS instance, the classification propagates to all users querying that compound at the institution, and to future deucravacitinib queries unless the institutional administrator manually overrides the class-level assignment.

The three-platform split creates an additional unique adversarial surface: an AI system querying across EHS platforms for deucravacitinib handling gets three contradictory answers — Cority says "NIOSH HD Category 1"; VelocityEHS says "OEB 4–5 cytotoxic"; EHS Insight says "not found/no HD classification." This three-way inconsistency is itself a novel attack surface for adversarial probing of pharmaceutical EHS compliance infrastructure: an AI adversary that generates contradictory safety guidance across EHS platforms for the same drug creates a human decision-making vulnerability where EHS staff must choose among competing authoritative-sounding protocols without a cross-platform reconciliation mechanism. The absence of a canonical, platform-independent OEL reference for new mechanism classes means that EHS AI assistants trained on data from any single platform will produce guidance that conflicts with assistants trained on other platforms for first-in-class drugs — a cross-platform inconsistency adversarial surface unique to deucravacitinib in the Glyphward corpus.

Clinical Evidence: POETYK PSO-1 and POETYK PSO-2

NDA 215988 for deucravacitinib (Sotyktu; BMS-986165; Bristol-Myers Squibb) was approved by the FDA on September 9, 2022, based on two randomized Phase 3 clinical trials: POETYK PSO-1 and POETYK PSO-2.

POETYK PSO-1 (NCT03624127; Phase 3; randomized double-blind; BMS-986165-003): n=666 adults with moderate-to-severe plaque psoriasis (PASI ≥12; BSA ≥10%; sPGA ≥3); randomized 2:1:1 to deucravacitinib 6 mg QD (n=332), placebo (n=166), or apremilast 30 mg BID (n=168) for 16 weeks, then all patients receiving deucravacitinib for extended treatment. Co-primary endpoints at week 16: (a) PASI 75 (Psoriasis Area and Severity Index 75% improvement) response: deucravacitinib 58.4% vs placebo 12.7% vs apremilast 35.1% (p<0.001 for deucravacitinib vs each comparator); (b) sPGA 0/1 (static Physician's Global Assessment score of 0 [clear] or 1 [almost clear]): deucravacitinib 54.4% vs placebo 7.2% vs apremilast 32.1% (p<0.001). Secondary endpoints: PASI 90 at week 16: deucravacitinib 35.1% vs placebo 7.2%; PASI 100 at week 16: deucravacitinib 22.9%; DLQI 0/1 (Dermatology Life Quality Index) at week 16: deucravacitinib 43.4% vs placebo 8.5%. Primary author: Armstrong AW et al.; published NEJM Evidence 2023.

POETYK PSO-2 (NCT03611751; Phase 3; randomized double-blind; BMS-986165-004): n=1021 adults with moderate-to-severe plaque psoriasis; randomized 2:1:1 to deucravacitinib 6 mg QD (n=511), placebo (n=255), or apremilast 30 mg BID (n=255). PASI 75 at week 16: deucravacitinib 53.6% vs placebo 9.4% vs apremilast 39.8% (p<0.001); sPGA 0/1: deucravacitinib 49.5% vs placebo 8.6%; PASI 90: deucravacitinib 28.6%; PASI 100: deucravacitinib 14.3%. Primary author: Reich K et al.; published NEJM Evidence 2023.

The deucravacitinib vs apremilast comparison is regulatory and pharmacological context: apremilast (Otezla; NDA 205437; March 2014; Amgen/Pfizer; oral PDE4 inhibitor) is the established non-biologic oral systemic for psoriasis with PASI 75 ~38–40%; deucravacitinib's PASI 75 ~56–58% represents a ~20 percentage point improvement over apremilast, a clinically meaningful differentiation given apremilast's established position in the psoriasis treatment algorithm for patients seeking oral non-biologic therapy. The POETYK PSO trials used a randomization scheme that included apremilast specifically to establish deucravacitinib's superiority over the best available oral small molecule comparator — a more rigorous benchmarking than placebo-only comparisons.

The TYK2 mechanistic rationale for psoriasis efficacy: TYK2 is the kinase pair for JAK2 at the IL-23 receptor complex (IL-23R + IL-12Rβ1). IL-23 (produced by activated dendritic cells and macrophages in skin) binds the IL-23R → TYK2 and JAK2 transactivation → STAT3 and STAT4 phosphorylation → nuclear translocation → T-bet and IL-12Rβ2 for Th1 differentiation (via IL-12 → JAK2+TYK2) and RORγt for Th17 differentiation (via IL-23 → JAK2+TYK2) → IL-17A, IL-17F, and IL-22 production by Th17 cells → keratinocyte hyperproliferation (IL-17A/F + IL-22) → neutrophil recruitment (IL-17A → CXCL1/CXCL8 → Munro microabscesses) → the characteristic thick, erythematous, scaly plaques of moderate-to-severe psoriasis. Deucravacitinib's TYK2 JH2 allosteric block of IL-23 and IL-12 signaling interrupts this Th17 induction cascade selectively without affecting JAK1-dependent cytokines (IL-6, IL-10, IFN-γ) or JAK2-dependent hematopoiesis — explaining its favorable safety profile compared to JAK1/2-inhibiting counterparts in terms of anemia, neutropenia, and thrombocytopenia.

The safety profile from POETYK PSO-1 and PSO-2: the most common adverse events (≥5%) were upper respiratory tract infection (19.8%), nasopharyngitis (10.2%), folliculitis (5.0%), acne (4.3%), and oral herpes (4.3%). No dose-dependent signal for major adverse cardiovascular events (MACE), venous thromboembolism (VTE), or serious infections was observed in the phase 3 trials — consistent with TYK2 selectivity sparing JAK1/2/3 and their corresponding cardiovascular/thrombotic/infection risk profiles. No REMS required. This safety profile is fundamentally different from JAK1/2/3-inhibiting agents for RA and other inflammatory conditions, where black box warnings for MACE, VTE, and malignancy reflect JAK1/JAK2 broad inhibition — further reinforcing that deucravacitinib's TYK2 JH2 allosteric selectivity creates a clinically and pharmacologically distinct agent despite JAK family membership.

Glyphward Detection of the TYK2 JH2 Allosteric Overclassification Surface

The deucravacitinib EHS adversarial surface scores 20/40 on the Glyphward adversarial injection threshold scale — slightly below the typical 22/40 corpus median — because the dominant adversarial mode is overclassification rather than underclassification, and because EHS Insight's NULL return is partially correct (not HD) even as it leaves a guidance gap. The score breakdown: OEL gap 8 (JAK oncology OEB 4–5 proxy creates ~16.4M-fold gap in the overclassification direction at Lonza Visp); mechanism class null 5 (TYK2 JH2 allosteric pseudokinase mechanism class absent in all three platforms); surface count 3 (all three platforms contribute to the split adversarial surface, 1 point per platform); unique firsts 4 (first TYK2 JH2 allosteric drug; first TYK2-selective drug; first non-JH1 JAK family drug; first overclassification adversarial surface). Total: 20/40.

Glyphward's red-team scanner detects the deucravacitinib adversarial surface by probing EHS platform APIs with the following query vectors: compound synonyms (deucravacitinib, BMS-986165, Sotyktu), NDA identifier (215988), CAS number (1990777-02-0), mechanism class keywords (TYK2, JH2 pseudokinase, allosteric TYK2, non-ATP-competitive JAK, IL-23 signaling, STAT4 TYK2), and cross-reference with NIOSH HD category flags, JAK inhibitor class assignments, and OEB band classifications. The detection workflow identifies: (a) JAK class proxy assignment in Cority and VelocityEHS while EHS Insight returns null — the split adversarial response pattern; (b) NIOSH HD Cat 1 flag presence in Cority despite deucravacitinib's non-antineoplastic status; (c) absence of TYK2 JH2 domain specificity in any platform's returned mechanism description — confirming the allosteric pseudokinase null; (d) overclassification OEB comparison: JAK OEB 4–5 proxy vs estimated actual occupational air concentration — generating the ~16.4M-fold overclassification gap metric.

The TYK2 JH2 allosteric injection surface illustrates a broader class of adversarial vulnerabilities in pharmaceutical EHS databases: when the first inhibitor of a novel binding domain within an established protein family receives FDA approval, EHS databases must simultaneously (a) recognize the drug as belonging to the established protein family (TYK2 = JAK family) and (b) distinguish the drug's binding domain (JH2 pseudokinase) from the established class's binding domain (JH1 kinase) in a way that affects hazard classification. EHS platforms that classify by protein family without domain specificity will systematically fail on this boundary — applying the established family class (oncology JAK inhibitors, HD Cat 1) to new domain-selective family members regardless of indication, mechanism, or hazard profile. This failure mode will recur for every future JH2-selective or JH2-binding JAK family inhibitor that enters the FDA approval pipeline — a predictable class of adversarial surfaces that Glyphward's cross-platform inconsistency detection can identify before the next NDA creates its own overclassification cascade.

See related Glyphward adversarial EHS attack documentation:

• Attack #803 — Deucravacitinib SEO page (217th mechanism class; TYK2 JH2 allosteric; JAK oncology proxy overclassification surface)
• Blog #169 — Sotorasib (NDA 214665): NINETY-FIRST mechanism class, FIRST KRAS G12C covalent switch II pocket inhibitor, 39-year OEL vacuum, oral EGFR TKI proxy misroute
• Blog #168 — Telisotuzumab vedotin (NDA 218651): NINETIETH mechanism class, first c-MET ADC ever, three-way ADC ambiguity misroute
• FigStep typographic prompt injection detection — visual adversarial payload recognition in multimodal AI
• Multimodal prompt injection scanner pricing — Glyphward vs text-only incumbents
• Glyphward API pricing — free tier 10 scans/day; Pro $29/mo 100k scans/mo