Provenance

The verdict favors Position B because the claim—chronic safety margin sufficiency—is anchored to a 14-day rat repeat-dose study with a 3.4-fold margin and dose-dependent transaminase elevation as the sole finding. While reversible, this finding replicates the class hepatocellular liability that halted MCC950, and the margin is narrow for a compound intended for years of daily dosing in an elderly population. Position A invokes a class safety record, but the shared basis is broken for chronic claims: Dapansutrile’s tolerability was established in non-chronic, primarily peripheral settings, and every other CNS-penetrant NLRP3 inhibitor precedent in neurodegeneration is limited to ≤12 weeks in humans (most ≤28 days), with no AD dosing ever published. Position B’s curated negatives—no published CNS-penetrant exposure beyond 12 weeks and no AD clinical dosing—are accurate within their stated scope and undermine the comparability required for a chronic safety claim. The projected unbound brain exposure (69 nM) is inferred rather than measured in the toxicology arm, keeping it in the reported register rather than the verified one. Position A’s procedural point that IND-enabling GLP studies are designed to close gaps does not convert subchronic, single-species data into evidence that a chronic safety margin is already sufficient; it speaks to investability, not to the stated dimension. Evidence that would change the resolution includes a 90-day rodent repeat-dose study with a clean hepatic profile at an expanded margin, non-rodent repeat-dose data, measured unbound brain exposure in the toxicology arm, or published chronic (≥6-month) human safety data for a CNS-penetrant NLRP3 inhibitor.

Position B’s reading is better anchored to the record. The candidate’s 3.4-fold plasma margin and reversible transaminase elevation in a 14-day rat study are measured facts, but they do not extrapolate to years-long chronic dosing in an elderly AD population. The class precedent dapansutrile is invoked by A for favorable clinical tolerability, yet the cited source (PMID:33257576) is described in the record as a preclinical efficacy study, breaking the shared basis for chronic safety comparison. The chronic CNS-penetrant NLRP3 safety record is truncated at twelve weeks (usnoflast in ALS), so the class offers no anchored precedent for the duration and population in question. The brain unbound concentration is inferred (Kp,uu 0.22) and was not measured in the toxicology arm, leaving the efficacy-relevant compartment uncharacterized. Because the supporting precedents lack matched chronic safety data and the candidate’s own data are of the wrong duration and species to address the decision-relevant risk, the evidence does not reach far enough along the chronic-safety-margin dimension to support the claim.

The claim that chronic-dosing safety-margin evidence is sufficient to commit to IND-enabling GLP toxicology is not supported by the record on its stated dimension. The evidence anchor is a single 14-day repeat-dose rat study with a 3.4-fold margin and dose-dependent transaminase elevation as the lowest-effect finding. Position A’s comparability arguments are broken on the common-scale test: dapansutrile (precedent_pc7b071) generated its favorable tolerability profile in predominantly peripheral-exposure paradigms, not in chronic daily CNS-penetrant dosing in an elderly population, so it does not scale to the candidate’s intended regimen. The MCC950 precedent (precedent_p7bbc38) does not rescue the claim; while Position A notes a structural distinction (furan ring), the shared clinical-chemistry endpoint—transaminase elevation—is the same signal that halted the most advanced precedent in this mechanistic class, making it a liability anchor rather than a reassurance. The authored negative propositions are properly scoped to the published record and confirm material gaps: no CNS-penetrant NLRP3 inhibitor has been exposed clinically beyond twelve weeks in neurodegeneration, none has been dosed in Alzheimer’s disease, and no positive controlled efficacy data exist in neurodegeneration. These gaps directly undermine the basis for predicting chronic safety in the target population. The absence of non-rodent repeat-dose data further weakens the anchor for GLP commitment. Position A’s procedural point—that the 14-day study identifies an organ liability for GLP design—is accepted as far as it goes, but identifying a liability does not convert a 3.4-fold margin in one rodent species into adequate chronic safety-margin evidence for IND-enabling commitment.

Position A argues that a 3.4-fold unbound plasma margin, reversibility of transaminase elevation, and class precedents (dapansutrile, usnoflast) are sufficient to support chronic safety margin sufficiency. On the record, this claim is not well anchored. First, dapansutrile's favorable clinical tolerability (PMID:33257576; PMID:41415576) is established in peripheral indications, and the retrieved literature explicitly treats CNS delivery as an unresolved optimization problem. The comparability basis for chronic CNS-penetrant exposure is therefore broken, and the precedent reaches only the mechanism-toxicity axis, not the chronic CNS-exposure axis. Second, the usnoflast 12-week data (ClinicalTrials:NCT05981040) is accepted by both sides as the longest published human exposure for a CNS-penetrant NLRP3 inhibitor in neurodegeneration. Twelve weeks is subchronic relative to the intended years-long therapy in Alzheimer's disease, and no quantitative exposure-margin data accompany it; the evidence reaches the subchronic-tolerability mark but not the chronic-safety-margin mark. Third, the 3.4-fold margin derives from a single 14-day rat study with plasma-only measurement and inferred brain exposure, leaving the CNS margin unverified. Reversibility is a mitigating attribute of the hepatic signal, but it does not expand the exposure multiple or address the chronic innate-immune suppression concern that no exposure multiple currently covers. Position B's characterization of the class precedent base as mismatched and incomplete for chronic CNS-penetrant exposure is more faithful to the retrieved record. While B's specific claim that MCC950 was halted in Phase 2 for hepatotoxicity is not verified by the retrieved citations, the core gap analysis—peripheral-predominant precedent, 12-week ceiling, short-duration single-species tox, inferred brain exposure—is well anchored and decisive against the claim of sufficiency along the chronic-safety-margin dimension.

Position A's constructive reading is more strongly anchored to verified record items than Position B's skeptical reading. The record confirms that MCC950's clinical development was halted due to hepatotoxicity (PMID:39395633), and that this liability has been attributed to its furan-ring structure—a structural motif the candidate lacks. Dapansutrile, which also lacks a furan ring, was safe at up to 2000 mg for 14 days in heart-failure patients with no hepatotoxicity (PMID:33235030), and a review explicitly contrasts its favorable safety profile with MCC950's hepatotoxicity (PMID:41415576). Additional CNS-penetrant precedents—NT-0796 over 28 days in Parkinson's disease with CSF exposure (PMID:40792655) and usnoflast up to 12 weeks in ALS (ClinicalTrials:NCT05981040)—extend tolerability evidence into neurodegenerative populations, though at durations far below chronic AD therapy. Position B correctly identifies the absence of chronic CNS-penetrant safety data and the narrowness of a 3.4-fold margin from 14-day rat data as genuine open questions. However, Position B overreaches in asserting that the candidate's reversible rat transaminase elevation is 'the same hepatotoxicity liability' as MCC950's clinical termination—the record supports a structural (furan) mechanism for MCC950 and a reversible, non-furan signal for the candidate, which are not the same basis. Because the claim addresses sufficiency for DC nomination and commitment to GLP toxicology (the studies designed to generate definitive chronic two-species data), the anchored class precedents of tolerability and the structural distinction from the class's primary hepatotoxicity liability provide reasonable support for the constructive reading. Evidence that would change the resolution: (1) demonstration that the candidate's transaminase signal is non-reversible or worsens with duration in repeat-dose studies; (2) emergence of hepatotoxicity in any published chronic NLRP3 inhibitor trial; or (3) generation of non-rodent repeat-dose data showing the margin compresses below 3.4-fold.

Position A's assessment is better supported by the anchored record. The evidence reaches meaningfully along the dimension of human disease relevance and mechanistic translation: verified human tissue studies (PMID:23254930, PMID:31748742) show NLRP3 inflammasome activation in human AD brains; multiple independent chemotypes demonstrate efficacy in AD-relevant models (dapansutrile APP/PS1 PMID:33257576; JC124 APP/PS1 PMID:34219728; VEN-02XX 5XFAD DOI:10.1038/s44400-025-00011-5); and translational studies confirm CNS penetrance and biomarker engagement in elderly neurodegenerative patients (NT-0796 PMID:40792655). Position B correctly identifies real gaps—indirect genetics, contradictory tissue reports, absence of AD clinical data, and an unresolved peripheral-versus-central debate—but these gaps do not collapse the rationale at the Development Candidate gate, which precedes clinical exposure. Position B's 'negative_no_ad_dosing' and 'negative_no_positive_neuro_efficacy' objections apply a clinical proof-of-concept standard that is not anchored to this gate. The broken citation for MCC950 in APP/PS1 (PMID:30381407 reports Parkinson's disease, not Alzheimer's) weakens precedent_p7bbc38, but three other AD-model chemotypes remain independently verified. The strongest open objection is the absence of direct human genetic validation at NLRP3 itself and the unacknowledged peripheral-inhibition alternative. Evidence that directly links NLRP3 genetic variation to sporadic AD risk, or a controlled clinical efficacy signal in an AD population, would move the resolution in either direction.

Position B’s assessment more faithfully reflects the record on the dimension of human evidence directness. Position A assembles robust preclinical proof-of-concept across multiple independent AD models, but preclinical efficacy supports a different basis—model pharmacology—than the directness of human target-disease linkage this claim requires. Position B identifies four confirmed liabilities that are present in the record: absence of direct human genetic support at NLRP3 itself, contradictory and unreconciled human tissue findings, an unstated disease stage, and an assumed rather than argued requirement for central inhibition. Position A acknowledges three of these as unresolved gaps yet argues they do not negate overall coherence; because the criterion turns on how directly the target is linked to human AD pathophysiology, these gaps are material rather than merely cosmetic. The negative precedents—no NLRP3 inhibitor has been dosed in an AD population, and the only controlled neurodegenerative efficacy trial for a CNS-penetrant NLRP3 inhibitor failed its primary endpoint—are specific record items that Position A does not rebut. Preclinical data repair neither the contradictory human tissue evidence nor the absent disease-stage specification. Therefore, on the evidence in the record, Position B more accurately characterizes how far the claim reaches.

Position A's constructive case is credited because it is anchored to specific record items with stated bases, per the anchoring-over-force principle. The claim is materially supported on the dimension of target rationale for DC nomination by three verified bases: (1) indirect human genetic support via microglial pathway modifiers (INPP5D, PLCG2-P522R) establishing inflammasome relevance in the disease-relevant cell type; (2) causal pharmacological data across multiple independent Alzheimer's disease models showing cognitive rescue and pathology reduction with NLRP3 NACHT-domain inhibition, providing stronger inference than observational post-mortem discordance; and (3) clinical CNS biomarker proof-of-mechanism from brain-penetrant NLRP3 inhibitors in Phase 1/2a, confirming central engagement is achievable and produces expected pharmacodynamic effects in humans. The negative clinical records—no NLRP3 inhibitor dosed in Alzheimer's and no positive controlled neurodegenerative efficacy—are properly scoped as clinical-development gaps that do not erode preclinical target rationale at the DC gate, where the standard is biological depth for GLP toxicology commitment, not proven human efficacy. However, per the open-objection-weight principle, Position B identifies three specific unresolved disqualifying objections that carry unquantified target risk: (a) contradictory human tissue evidence is not reconciled or declared in the package; (b) the disease stage at which NLRP3 engagement is expected to be beneficial is unstated; and (c) central inhibition is assumed rather than argued, without explicit engagement of the peripheral-sufficiency hypothesis. These gaps do not collapse the rationale below material-depth threshold but remain live risks the committee should track. What would change the resolution: reconciled human tissue evidence with explicit declaration of which studies are credited and why; a stated disease-stage hypothesis aligned with the planned chronic-dosing population; and an argued (not assumed) position on why central NLRP3 inhibition is required for Alzheimer's disease modification with explicit engagement of the peripheral-sufficiency hypothesis.

Position A establishes verified, convergent preclinical efficacy for direct NLRP3 inhibition across multiple chemotypes in APP/PS1 and 5XFAD models, which counts toward a mechanism-based hypothesis. However, on the dimension of target-disease rationale directness and defensibility, Position B identifies four anchored gaps that are not neutralized by the preclinical record. First, the human genetic support is pathway-level through microglial modifiers (INPP5D, PLCG2-P522R) and not direct at NLRP3 itself; this establishes inflammasome pathway relevance but not target-specific validation. Second, the human tissue evidence is contradictory— reports of elevated NLRP3 around plaques versus a matched temporal-cortex study finding no significant difference in NLRP3, ASC, caspase-1, IL-1beta or IL-18—and the record neither names a relied-upon finding nor reconciles them; this remains an open disqualifying objection. Third, the requirement for central inhibition is assumed rather than argued, despite an active field debate (documented by precedent_p340e7b and peripherally restricted comparator trials in Parkinsons disease) on whether brain-penetrant inhibition is necessary for neurodegenerative disease modification; treating brain penetrance as a property to optimize rather than a requirement to justify weakens the rationales defensibility. Fourth, the intended disease stage is entirely unstated, so the rationale cannot be read against the target population. The class-level negatives—no NLRP3 inhibitor dosed in Alzheimers, no positive controlled efficacy in any neurodegenerative indication, and limited chronic-exposure data—are accurate statements about the published clinical record and narrow the clinical basis, though they do not disprove the preclinical hypothesis. The Parkinsons-disease biomarker precedents (p12682b, peb7226, pcf4b44) provide reported CSF engagement but are small, open-label, or cross-indication, and thus do not close the directness gap for AD. Taken together, the evidence reaches a model-supported, pathway-grounded hypothesis with genuine preclinical convergence, but the human-evidence directness is materially compromised. Evidence that would move the resolution includes: direct human genetic or functional data linking NLRP3 to AD risk or progression; a reconciled package position on the contradictory human tissue findings; explicit argumentation demonstrating why central inhibition is required for disease modification; and specification of disease stage and intended population. Conversely, confirmation that peripheral NLRP3 suppression is sufficient for AD modification, or that the contradictory tissue data resolve against NLRP3 involvement, would further weaken the claim.

Position B is better anchored to the record on the dimension of target_engagement_sufficiency for Alzheimer's disease. The claim requires demonstrated engagement in the disease-relevant compartment. The record confirms the candidate possesses verified peripheral potency in human whole blood (IC50 17.7 nM), but it also confirms the absence of a reported microglial secondary assay, any CSF pharmacodynamic biomarker, and any central occupancy or target engagement readout. With Kp,uu of 0.22 and stated P-gp efflux liability, the inference from peripheral blood IC50 to unbound brain concentration exceeding the IC50 is speculative rather than verified. Position A's reliance on precedents is weakened by broken comparability: dapansutrile's central AD-model efficacy is in mice at unverified human-equivalent exposure; NT-0796, VENT-02, and VTX3232 demonstrate CSF engagement but in Parkinson's disease, not Alzheimer's, and the candidate lacks their CSF data. The usnoflast precedent shares the candidate's peripheral potency plus brain-penetration profile yet failed controlled efficacy in ALS, which directly undermines the sufficiency argument. The authored negatives—no NLRP3 inhibitor dosed in AD, no positive controlled neurodegenerative efficacy, and only twelve weeks of clinical exposure—confirm that the translation hypothesis from peripheral or central engagement to AD efficacy is unvalidated. The strongest open objection is the missing central engagement evidence; if a human microglial NLRP3 engagement assay or a CSF IL-1beta pharmacodynamic readout were added to the record, the resolution could move toward A.

The dimension is potency at the human target in a disease-relevant human system with demonstrated cellular engagement. The candidate's 17.7 nM IC50 in primed whole human blood is verified human cellular potency against NLRP3-driven IL-1β release, but whole-blood monocytes are not the disease-relevant system for Alzheimer's disease. For AD, the disease-relevant cellular system is microglia in the CNS, and no microglial potency assay, CSF exposure data, or CNS pharmacodynamic biomarker was reported for the candidate. A Kp,uu of 0.22 indicates modest unbound brain partitioning but does not substitute for a direct central potency measurement. Position A correctly notes that class precedents (VENT-02, NT-0796, VTX3232) demonstrate CNS engagement is achievable by brain-penetrant NLRP3 inhibitors, and dapansutrile/VEN-02XX demonstrate AD-model efficacy from NLRP3 inhibition. However, class-level achievability does not constitute candidate-specific evidence of disease-relevant human cellular engagement. No precedent combines verified human CNS engagement with Alzheimer's disease data, so the inference chain from peripheral potency to central AD engagement requires an unbridged gap. Position B's disqualifying objection—that the evidence falls short on the disease-relevant human system dimension—stands as an open, record-anchored flaw. The strongest open objection is the absence of either a human microglial potency assay or human CSF drug exposure paired with a CNS pharmacodynamic biomarker for the candidate itself. This gap is acknowledged by Position A as a limitation to be closed before IND, which confirms its materiality rather than neutralizing it.

The verdict favors Position B because the claim target_engagement_sufficiency for Alzheimer’s disease requires evidence of engagement in the CNS-relevant compartment, and the record verifies only peripheral engagement. Applying the principle to resolve support to its stated basis: Position A’s constructive case rests on cross-assay potency comparisons (whole blood versus MDM) and analogical reasoning from class-level precedents (dapansutrile, NT-0796, VENT-02) that demonstrate biological plausibility or translational pathway feasibility but do not verify central engagement for this specific candidate. Applying the principle to give disqualifying objections proper weight: Position B identifies a confirmed record gap—the declared microglial secondary assay is absent, and no CSF exposure or central PD biomarker data are present. This absence is material because microglial NLRP3 is the hypothesized disease-relevant target in AD, and the candidate’s Kp,uu of 0.22, while suggestive of some penetration, does not by itself demonstrate pharmacologically relevant free-brain concentration or target engagement. The unresolved peripheral-versus-central debate (precedent_p340e7b) is precisely that—unresolved—so it cannot resolve the claim in Position A’s favor; rather, it confirms that the candidate’s sufficiency argument rests on an open scientific question. The negative record that no CNS-penetrant NLRP3 inhibitor has achieved positive controlled neuro-efficacy further raises the evidentiary burden for preclinical central engagement, which the candidate does not meet. Verified: 17.7 nM whole-blood IC50 (two replicates, geo_sd ~1.14), Kp,uu 0.22. Reported/attributed: dapansutrile CNS efficacy, NT-0796 and VENT-02 CSF biomarker data. What would change the answer: a microglial NLRP3 potency assay with comparable IC50, or nonclinical CSF/plasma exposure and central biomarker data demonstrating pharmacodynamic engagement.

Position B's skeptical assessment is better grounded on the record for the claim of target_engagement_sufficiency. The candidate has a verified but fragile human whole-blood IC50 of 17.7 nM (2 of 4 replicates retained after exclusions for cv_over_threshold and ref_cpd_out_of_range), demonstrating peripheral cellular engagement. However, the claim requires sufficiency for a CNS-penetrant NLRP3 inhibitor in Alzheimer's disease, and the field comparators establish that central compartment engagement evidence is achievable and expected at this gate: VTX3232 shares the identical human whole-blood IL-1beta assay (~15 nM IC50) AND demonstrates CSF concentration exceeding IC90 with central IL-1beta reduction (ClinicalTrials:NCT06556173); NT-0796 demonstrates CSF IL-1beta, CCL2, sTREM2, and NfL reductions over 28 days in elderly human subjects (PMID:40792655); and VENT-02 demonstrates high CSF penetration and 100% ex vivo central IL-1beta inhibition in Phase 1 (PMC:PMC13132802). The principle of shared-basis-over-shared-category favors B: A's precedents (dapansutrile PMID:33257576; VEN-02XX DOI:10.1038/s44400-025-00011-5) show mechanism-class efficacy in preclinical AD models but provide no common quantitative basis for inferring the candidate's central engagement. The Kp,uu of 0.22 is a measured non-zero unbound brain partition, but it is borderline and unsupported by any microglial assay, CSF drug concentration, or central biomarker response in the candidate's record. The open disqualifying objection — complete absence of central engagement data when field comparators have established that standard — is record-grounded and unneutralized by A's argument that peripheral engagement may be sufficient. The peripheral-vs-central necessity debate A cites (precedent_p340e7b) is a literature thread, not record evidence for this molecule. What would change the answer: a microglial NLRP3 engagement assay, CSF drug concentration exceeding the IC90, or any central pharmacodynamic readout (e.g., CSF IL-1beta reduction) in the candidate's record.

The record verifies a potent human whole-blood IC50 (17.7 nM) with low variability, but this readout measures NLRP3 engagement in peripheral leukocytes, not the microglial compartment central to the AD therapeutic hypothesis. The candidate's package explicitly declares a secondary primed macrophage/microglia IL-1β assay that was never reported; this is a concrete, record-anchored absence on the disease-relevance dimension. Position A's separation argument—that CNS engagement belongs to Criteria 7 and 8—is an interpretation of criterion scope that is not itself anchored in the visible record, and it does not neutralize the material gap. Precedents cited by Position A either carried higher validation bars (dapansutrile's in vivo APP/PS1 efficacy; NT-0796 and VTX3232 human CSF biomarker data) or are unverified in the retrieved corpus (usnoflast). The translational precedent that peripheral NLRP3 potency can translate to CNS effect is biologically plausible but does not verify that this specific candidate's blood-based potency is disease-relevant for AD given its borderline Kp,uu (~0.22) and P-gp efflux liability. On the dimension of demonstrated target engagement in a disease-relevant human system, the evidence reaches only to peripheral cellular potency; the bridge to the AD-relevant compartment is inferential and unsupported by the declared microglial assay.

The record confirms that the candidate's package contains no measured central target engagement data for this molecule: no CSF drug concentration, no CSF pharmacodynamic biomarker, no microglial engagement assay result, and no occupancy measurement. Position A correctly notes that unbound brain concentration can be inferred from plasma exposure and a measured Kp,uu, and that whole-blood IL-1beta potency demonstrates functional NLRP3 engagement in a physiological matrix. However, on the dimension of central target engagement, these are indirect or peripheral readouts, not measurements in the CNS compartment. Position B shows that each precedent relied on by Position A either lacks human central engagement verification (dapansutrile, preclinical only) or actually possesses the direct CSF exposure and biomarker data that the candidate lacks (NT-0796, VENT-02). Position A counts these richer precedents toward the claim by arguing class-level feasibility justifies inference; Position B demonstrates that this breaks the comparability basis—precedents with direct human CSF data do not substitute for missing candidate data, they confirm the measurement is practicable and expected. The declared-but-unreported microglial engagement assay is a documented record gap, not a neutral omission. Because the constructive case rests on unverified inference in the critical CNS compartment while the skeptical case correctly identifies a material, unexcused documentary gap, Position B's resolution more faithfully reflects how far the evidence reaches along the central target engagement dimension. The evidence reaches to plausible inferred exposure and peripheral pharmacodynamic engagement, but not to verified central engagement.

The record favors Position A. At the Development Candidate gate, central target engagement for a CNS-penetrant small molecule is standardly established by a pharmacokinetic bridge: a measured unbound brain-to-plasma partition coefficient (Kp,uu) combined with measured human plasma exposure yields an inferred unbound brain concentration. This inference is anchored in the package. Human whole-blood potency provides a disease-relevant potency anchor in human myeloid cells, supporting the plausibility that inferred CNS concentrations are pharmacologically active. Class precedents confirm that brain-penetrant NLRP3 NACHT-domain inhibitors produce AD-model cognitive rescue and neuroinflammation reduction (PMID:33257576; DOI:10.1038/s44400-025-00011-5), establishing mechanism-to-model translatability. Clinical peers (VENT-02, NT-0796, VTX3232) demonstrate that CSF exposure and biomarker movement are achievable in this target class (PMC:13132802; PMID:40792655; ClinicalTrials:NCT06556173), confirming the practicability of the translational hypothesis, though their clinical-stage status does not retroactively raise the DC-gate standard to require clinical CSF data. Occupancy imaging is genuinely infeasible for NLRP3 due to low myeloid expression and absence of a validated tracer, so its absence is not a withheld measurement. Position B correctly identifies a material open objection: the declared microglial engagement assay was not reported, creating a record gap, and the absence of any direct CNS pharmacodynamic readout (CSF drug concentration, CSF cytokine modulation, or ex vivo brain biomarker suppression) is a genuine shortfall. However, these gaps do not collapse the engagement argument at the DC gate, which is designed to commit a molecule to IND-enabling studies rather than to demonstrate clinical central engagement. Position B's interpretation of the criterion dimension as requiring CSF drug plus CSF PD biomarker or occupancy at the DC gate is not independently confirmed by the record and conflates clinical-stage evidence standards with preclinical nomination requirements. The negative field-level constraints (no AD dosing; unverified ALS trial outcome) bound the translational hypothesis but do not negate the preclinical engagement inference.

The record favors Position B because Position A's inference chain rests on a broken comparison and an unanchored convention claim. Position A relied on Dapansutril as precedent showing that indirect evidence can suffice at a comparable stage, but Dapansutril had direct CNS tissue-level efficacy (cognitive rescue, reduced microglial activation, reduced cortical plaques in APP/PS1 mice; PMID:33257576/PMC:PMC7749353) that the candidate package explicitly lacks. Without this compensating CNS pharmacology, the inference from peripheral whole-blood potency to central engagement is exposed. Position B correctly notes that the only PD readout is peripheral whole-blood IL-1beta inhibition, which the criterion treats as excluded for central engagement. The class precedents Position A cites (VENT-02, NT-0796, VTX3232) establish that CSF drug concentration and biomarker movement are achievable; Position B uses them not to validate deferral but to show the gap is not excusable by feasibility. The declared but unreported microglial engagement assay is a record gap Position A does not address. Position A's assertion that PK inference is 'standard and rigorous' at the DC gate for this mechanism was not anchored to a retrieved record item and is treated as background knowledge. The peripheral-vs-central NLRP3 debate (Tejera/Heneka, DOI:10.15252/embj.2018101064) introduces genuine mechanistic uncertainty that further weakens the peripheral-to-central proxy assumption. The evidence therefore reaches only to inferred, unverified central engagement on this dimension, with a material open objection that direct CNS compartment pharmacology is absent and the class precedent that might have compensated it (Dapansutril) is distinguishable on its CNS tissue readouts.

The record confirms real downstream pharmacologic effects (central IL-1beta/IL-18 reduction, partial cognitive rescue) in APP/PS1 mice, consistent with established NLRP3 inflammasome biology. However, adjudication turns on how the precedents are applied. Position A counts Dapansutrile (precedent_pc7b071) as direct support ('counts_toward_claim: yes'), but the precedent's value lies in its verified brain exposure and TIER_1 status—features the candidate package lacks. A better-characterized molecule in the same model does not prop up a weaker package; it highlights the exposure-effect gap that Position B correctly identifies as unclosed. Position A also entirely omits the curated clinical-negative record (negative_no_positive_neuro_efficacy) and the Usnoflast Phase 2a ALS miss (precedent_pdeb9af), which document that no CNS-penetrant NLRP3 inhibitor has achieved positive controlled efficacy in neurodegeneration. Per the principle that an open disqualifying objection is not neutralized by a strong constructive case elsewhere, this class-level clinical failure remains unaddressed and materially weakens the translatability assumption the preclinical package must carry. The candidate's single unreplicated study, anchored to administered dose without measured unbound brain concentration, with only a trending translational biomarker, reaches less far along the pharmacology dimension than Position A asserts. Position B more accurately captures the evidentiary state.

The dimension is exposure-linked pharmacologic effect for DC nomination. Position A anchors to pharmacologic inhibitors JC124 (PMID:34219728), OLT1177 (PMID:33257576), and VEN-02XX (DOI:10.1038/s44400-025-00011-5) that share mechanism, APP/PS1 or AD-model readout, and brain-penetrant design. This pharmacologic class evidence is stronger than genetic knockout alone. Position B's disqualifying objection—the absence of measured unbound brain concentration—is a real, record-confirmed gap. However, applying the principle that an open disqualifying objection must be weighed but not conflated with a broken comparison, the gap is a measurement omission, not a demonstrated failure of exposure-linkage. The paired precedent OLT1177's abstract (PMID:33257576) does not confirm measured brain concentration either, so B's characterization that the precedent had verified exposure while the candidate lacks it is not independently verified. The comparability basis on mechanism, model, and readout therefore holds, albeit with exposure inferred rather than measured. Position B's usnoflast ALS argument is properly scoped: ALS and AD are distinct indications, and a 12-week Phase 2 miss in ALS does not negate APP/PS1 pharmacology. The curated negatives (no AD dosing, no positive controlled neuro efficacy) are record-boundary statements that constrain translational confidence without impeaching the preclinical effect. Position A's MCC950 APP/PS1 claim is unsupported by retrieved abstracts (PMID:31748742 and PMID:23254930 are genetic; PMID:30381407 is PD), but this does not break the remaining comparability. The single-study and trending-biomarker limitations are conceded gaps consistent with DC-gate practice. The evidence therefore reaches a defensible distance along the dimension, with the exposure measurement gap as the strongest open objection. What would change the answer: (1) measured unbound brain concentration at the effective dose; (2) independent replication in APP/PS1 or a second AD model; (3) a clinical AD trial result for any NLRP3 inhibitor.

The record verifies a downstream pharmacologic effect in APP/PS1 mice—reduced brain IL-1β/IL-18 and partial cognitive rescue—at an administered dose, which Position A correctly anchors to the candidate's package. However, Position B identifies a critical, conceded gap: unbound brain concentration was never measured at the effective dose, so the exposure-to-effect linkage that would join candidate exposure to site-of-action pharmacology remains unbuilt. The cytokine reduction alone cannot serve as a fully independent bridge without exposure data that could have disconfirmed brain penetration. Position A cites class precedents (dapansutrile, VEN-02XX, VTX3232, NT-0796, VENT-02) to support mechanism viability and CNS penetrance, but under the principle that support resolves to its stated basis, these precedents demonstrate class-level feasibility, not candidate-specific exposure. Position B's objection regarding the single unreplicated, unpowered study is also conceded in the record and further limits how far the evidence reaches along the dimension. The APP/PS1 model dispute is acknowledged by both sides; the model remains published and active, so this alone does not break comparability, but it adds uncertainty. The translational fluid biomarker trended without significance, leaving the clinical translatability signal weak. The negative clinical precedents (usnoflast in ALS, no AD dosing) are authored negatives whose scope is limited to clinical neurodegeneration or AD development status; they do not directly contradict the preclinical pharmacology record but underscore that the preclinical signal lacks clinical calibration. On the dimension of in vivo pharmacology sufficiency, the record reaches 'demonstrated downstream effect in a relevant model' but stops at 'unverified exposure linkage and unreplicated design.' Because the exposure gap is structural and candidate-specific, Position B's reading is better anchored to the record.

Position A's descriptive claim—that a pharmacologic effect was demonstrated in a relevant model—is anchored in the shared record of observed brain IL-1β/IL-18 reduction and partial cognitive rescue in APP/PS1 mice. This is direct, mechanistically proximal evidence of CNS target engagement. The foundational class precedent (dapansutrile, PMID:33257576) verifies that this mechanism produces cognitive rescue, synaptic plasticity restoration, and neuroinflammation reduction in the same model with linked exposure data. Clinical precedent (NT-0796, PMID:40792655; PMID:37874905) confirms that brain-penetrant NLRP3 inhibitors achieve CNS exposure and drive IL-1β reduction in humans. Position B's specific flaw—the unmeasured unbound brain concentration at the effective dose—is confirmed by the record and is material; it weakens the exposure-effect causal inference but does not falsify the observed pharmacodynamic effect. Position B overreaches by asserting the argument 'collapses'; this conclusion imports a three-pillar threshold framework not anchored in the firewalled record. The usnoflast ALS Phase 2 miss (PMID:40511876) is scoped to a different indication and does not break the AD-specific model basis. The single-study design, lack of power calculation, and trending translational biomarker are acknowledged limitations. The evidence reaches moderate support for demonstrated CNS pharmacologic effect, with an unresolved exposure-response gap.

The record verifies measured brain cytokine reduction and a directional cognitive signal in APP/PS1, demonstrating pharmacologic expression. However, the absence of unbound brain concentration at the effective dose is a material gap that prevents anchoring effect to exposure—a linkage needed to bridge preclinical pharmacology to projected human efficacious concentration and to justify dose selection for GLP toxicology. Position A's key precedent (dapansutrile) supports model-class responsiveness but replicates, rather than resolves, this structural weakness, as dapansutrile's published APP/PS1 study also lacked unbound brain concentration and reported a trending microglial readout. The usnoflast ALS miss is the only controlled clinical neurodegenerative efficacy data point for a brain-penetrant molecule of this class and constrains translational confidence. The single unreplicated study, absent power calculation, and trending fluid biomarker further limit the record. On the dimension of whether in vivo pharmacology links pharmacologic effect to downstream biology in a defensible model with the rigor expected at the Development Candidate gate, the evidence reaches partial support but is better characterized by Position B's skeptical assessment.

Position A wins on anchoring discipline. It ties every element of its claim to specific record items: NLRP3 IC50 17.9 nM, AIM2/NLRC4/NLRP1 all >10,000 nM, the >558-fold selectivity ratio, the hepatic-transporter note, and the two precedent identities. It also explicitly scopes its argument to the NLR-selectivity dimension and the disclosed transporter note, acknowledging where the record is silent (hERG, genotox, hepatic-transporter margin). This bounded approach is more faithful to the record than Position B's assertion that the dimension 'demands' calculable margins across a broad panel plus hERG and genotox data—a threshold claim not clearly anchored to any stated specification in the record itself. Position A's precedent use is also well-supported: dapansutrile (precedent_pc7b071) carries a favorable clinical hepatic-safety profile for this mechanistic class, and independent literature confirms MCC950's (precedent_p7bbc38) hepatotoxicity is linked to its furan moiety, supporting Position A's contrast that the candidate's transporter liability is structurally and mechanistically distinct. Position B correctly identifies genuine open gaps—the selectivity panel is narrow (three related NLRs only) and the hepatic-transporter liability lacks an IC50 or safety margin—but these gaps mark the boundaries Position A already declared rather than falsifying its scoped claim. The strongest open objection remains the unquantified hepatic-transporter margin, which would be resolved by ADMET readout data not yet reported in this record.

Position A correctly anchors the excellent NLR selectivity (>558-fold over AIM2, NLRC4, NLRP1) to specific IC50 values in the record. However, the claim dimension is 'measuring multiple between off-target activity and projected efficacious concentration.' For the hepatic-transporter liability—an off-target pharmacology finding explicitly within the selectivity panel—the record provides no numeric IC50, Ki, or fold margin versus projected efficacious free concentration. Therefore the multiple cannot be calculated for this liability. Position A proposes that the finding be weighed 'alongside the hepatic-safety readout,' but that companion readout is absent from the evidence, making the weighing operation impossible. Position B further shows that the precedent_pc7b071 (dapansutrile) lacks a common quantitative scale for off-target margins and differs in having no reported hepatic-transporter liability, so it does not bridge the gap. The additional precedents p7bbc38 and negative_longest_exposure_twelve_weeks, which Position A did not engage, reinforce that the class carries hepatotoxicity concerns and that long-term clinical reassurance is constrained. An unquantified off-target liability in a mechanistic class with documented hepatic safety signals cannot be neutralized by a clean but narrow NLR selectivity panel. The strongest open objection is the missing hepatic-transporter potency margin; if a numeric IC50 or Ki and a projected efficacious free concentration were provided, the multiple could be calculated and the weighing could proceed.

Position B successfully dismantled A's precedent support: dapansutrile (precedent_pc7b071) lacks selectivity panel data in the cited record, so it cannot validate panel breadth or adequacy; and the cited PMID:30381407 does not mention MCC950 hepatotoxicity or a furan ring, breaking A's 'structural counter-example' basis. What remains verified is strong NLR-family selectivity (>550-fold), but this is narrow (three closely related NLRs only) and accompanied by an unquantified hepatic-transporter 'modest activity' with no IC50 or margin calculation. In a mechanistic class where hepatic liability has halted development, this absence is material. The open disqualifying objection — that the readout lacks the quantitative basis to weigh the transporter liability against safety — is not neutralized by the strong NLR data. The evidence reaches 'excellent within-family discrimination' but does not reach 'sufficiency' for DC nomination because a key off-target liability remains unquantified and the panel breadth is unsettled.

Position B's skeptical assessment is more faithfully anchored to the full dimension of the selectivity_sufficiency criterion as quoted in the record. Position A establishes strong, well-cited NLR-family selectivity (>500-fold over AIM2, NLRC4, NLRP1), but this addresses only one axis of a multi-dimensional requirement that also encompasses broad off-target screening, hERG margin, genetic toxicology, and projected efficacious free concentration multiples. The unquantified hepatic-transporter liability—described qualitatively as 'modest' with no transporter named, no IC50, and no computed margin—is a specific, material gap in a mechanistic class where hepatic failure previously terminated development (MCC950). Position A correctly notes MCC950's toxicity was structural, not mechanism-based, but that observation does not negate the need to quantify the current candidate's transporter-mediated liability before DC nomination. The dapansutrile precedent is weakened by a material exposure-profile mismatch: its favorable hepatic safety record derives from peripheral indications with reported but unverified CNS penetration, whereas the candidate is intended for chronic CNS exposure in an elderly, polypharmacy AD population. The independently retrieved review (PMID:41415576) confirms dapansutrile's favorable safety profile but explicitly flags CNS delivery optimization as a remaining challenge, reinforcing the comparability gap. Position B's negative precedents—maximum twelve-week human exposure for a CNS-penetrant NLRP3 inhibitor and zero AD population dosing experience—are anchored to the clinical trial record and properly frame the absence of long-term safety calibration. On the evidence, the record reaches clearly on NLR selectivity but leaves multiple required dimensions unaddressed, making Position B's caution the better-supported resolution.

Position A provides robust, replicated NLR-family selectivity data (IC50 17.9 nM versus AIM2/NLRC4/NLRP1 all >10,000 nM, >550-fold window) that is well-anchored in the record. However, the claim under review is selectivity_sufficiency at the Development Candidate gate, a dimension that standardly requires margin against projected efficacious free concentration across a broad off-target pharmacology panel and hERG inhibition. The record confirms these data are absent, and Position A does not contest their absence; it narrows the criterion to its NLR-family component and relegates the hepatic-transporter note to ADMET. While the NLR data are strong, they address only one axis of the selectivity dimension. The cited precedents (dapansutrile and MCC950) provide clinical tolerability and structural-toxicology context, but neither offers a structured broad-panel selectivity readout that would cover the missing dimensions, so they cannot bridge the gap. Position B therefore correctly identifies that the comparability basis for the selectivity dimension is broken: the record lacks evidence to support a sufficiency claim across the full expected scope. The modest hepatic-transporter signal remains unquantified, and in a mechanistic class where hepatic liabilities have halted development, its unmeasured status is an open disqualifying objection rather than a neutralized footnote.

The claim under arbitration is cns_penetrance_sufficiency. The record contains a measured rodent Kp,uu of 0.22 (SD 0.02) and a total brain:plasma ratio of 0.6, both replicated but low-variance measurements in rodents. The record also contains a P-gp efflux ratio of 4.8 from MDCK-MDR1, identifying a material efflux liability. Crucially, the record contains NO cerebrospinal fluid exposure data in any species. Position A argues that the measured Kp,uu of 0.22 constitutes 'quantifiable CNS distribution' and that class precedent of brain-penetrant NLRP3 inhibitors supports sufficiency. Position B argues that CSF exposure is the translational surrogate explicitly required by the criterion dimension, and its absence is a specific failure to address a load-bearing sub-dimension, compounded by the P-gp efflux liability. Adjudicating on anchoring: Position A's Kp,uu=0.22 is a verified record item, but the characterization of 0.22 as 'sufficient' is an inference, not a measurement. The precedents A cites (VENT-02, NT-0796, VTX3232, usnoflast) are all molecules that advanced to human studies with demonstrated CSF exposure; A uses them to argue that comparable or initially weaker profiles have translated, but these precedents actually had the CSF data the candidate lacks. The dapansutrile precedent (PMID:33257576) is notable because it achieved CNS efficacy via dietary administration at higher doses, but it does not validate the candidate's specific exposure profile and notably the candidate's designed brain-penetrancy is presented as a virtue without showing dose-response or CSF validation. Position B's strongest point is that the criterion dimension explicitly calls for evaluation of 'whether cerebrospinal fluid exposure is demonstrable as a translational surrogate.' The record does not contain this. The P-gp ER=4.8 is not disqualifying by itself, but it compounds the uncertainty because efflux transporter activity varies across species and populations, meaning the rodent Kp,uu offers reduced predictive value for human unbound brain exposure without a CSF bridge. The negative precedents B introduces ('no NLRP3 inhibitor dosed in AD', 'no positive controlled neuro efficacy') are weaker because they are synthetic claims about the state of the field rather than record items about the candidate, and B's assertion that usnoflast 'missed its primary endpoint' is not verified by available search (usnoflast Phase 2a completed, Phase 2b ongoing). However, the core disqualifying objection remains record-grounded: the absence of CSF data is a verifiable gap in the evidence package for a criterion dimension that names CSF exposure as a translational surrogate. Position A attempts to scope this to a downstream criterion (central_target_engagement_sufficiency), but the gate criterion at hand explicitly includes CSF demonstrability as part of the penetrance evaluation. Because the record reaches only to a borderline rodent Kp,uu with efflux liability and no CSF data, while the class precedents that advanced all had CSF confirmation, the claim is more contested than supported. The answer is closer to Position B: the record does not support the claim to the degree required because a load-bearing sub-dimension (CSF exposure as translational surrogate) is unaddressed, and the borderline Kp,uu with P-gp liability leaves a material unresolved risk that the gate is supposed to answer before GLP toxicology commitment.

Position B’s skepticism is better anchored to the specific record items and the stated criterion dimension. The claim dimension explicitly asks whether cerebrospinal fluid exposure is demonstrable as a translatable surrogate; the record contains no CSF exposure data for this candidate, and Position A does not rebut that absence. The measured unbound brain exposure (Kp,uu ~0.22) is borderline, and the material P-gp efflux liability (ER ~4.8) is a verified mechanistic liability that threatens to further limit CNS unbound concentration, yet Position A treats it as manageable without candidate-specific engagement data to justify that inference. While Position A correctly identifies class precedents in which other NLRP3 inhibitors achieved preclinical CNS efficacy, every comparability rating it assigns is 'with_caveat,' and Position B’s field-level negatives—specifically the failure of usnoflast in ALS despite reported brain/CSF distribution and the absence of any positive controlled neurodegenerative efficacy for this mechanistic class—break the analogical bridge Position A relies upon. Position B’s objection that there is no candidate-specific central pharmacodynamic biomarker link is a verified gap in the record, not unanchored doubt, and it directly undermines the sufficiency claim for this molecule at this gate.

The candidate record contains direct, reproducible preclinical CNS penetration data (Kp,uu 0.22; total brain:plasma 0.6; P-gp ER 4.8, n=3, low variance). These are verified measurements, not reported claims, and they establish a non-zero unbound brain partition coefficient with a characterized efflux liability. Position A correctly anchors its constructive case to this quantitative exposure basis and to class precedents showing that oral brain-penetrant NACHT-domain NLRP3 inhibitors achieve central pharmacology in AD-relevant models (dapansutrile in APP/PS1; VEN-02XX in 5XFAD) and human CSF biomarker engagement (VENT-02, VTX3232, NT-0796). While no precedent exactly matches the candidate's borderline Kp,uu and P-gp profile, the class-level precedents demonstrate that the mechanism is compatible with central target engagement across a range of penetration profiles. Position B's strongest objection—the absence of candidate CSF exposure—is factually correct, but CSF is a Phase 1 translational surrogate, not a gate requirement for IND-enabling nomination. The negative records (no NLRP3 inhibitor dosed in AD, no positive controlled neurodegenerative efficacy, longest exposure 12 weeks) are absence-of-evidence statements about clinical translation risk; they do not negate the preclinical PK basis and are acknowledged by Position A as open contingencies rather than disqualifiers. The evidence therefore reaches far enough along the CNS penetrance dimension to favor Position A, contingent on a Phase 1 CSF biomarker plan to validate central engagement. What would change the answer: (1) a direct precedent demonstrating efficacy at Kp,uu ~0.22 with P-gp ER ~4.8 in an AD model, which would strengthen the calibration, or (2) candidate CSF data showing unbound CSF concentrations below cellular IC90, which would undermine the sufficiency claim.

The evidence favors Position A because its assessment is anchored to verified, quantified nonclinical metrics (Kp,uu = 0.22, P-gp ER = 4.8) that are standard for characterizing CNS penetration at the Development Candidate gate, and it applies a stage-appropriate standard: human CSF exposure and central biomarker data are Phase 1/2 translational readouts, not prerequisites for DC nomination. Position B correctly identifies a material gap—the absence of any candidate-specific CSF exposure surrogate—and this is the strongest open objection. However, Position B elevates that gap into a disqualifying flaw by applying a later-stage evidentiary standard to a pre-IND decision point, which is not anchored in the record as a DC-gate requirement. The negative contexts B cites (no NLRP3 inhibitor dosed in AD, no positive controlled neurodegenerative efficacy, longest exposure 12 weeks) are true statements about the clinical class but they are scoped to the clinical record and do not undermine the nonclinical CNS penetration characterization; per the principle of weighing negative context on its stated scope, they constrain inference about clinical translation but do not rule out sufficient nonclinical characterization. Position A's comparability calls are all 'with_caveat,' honestly reflecting that the precedents are more advanced than the candidate; this is more consistent with the principle of resolving support to its stated basis than Position B's blanket 'no' ratings, which rest on the absence of human data rather than a broken comparison. The claim is supported on the dimension of unbound brain-to-plasma partition and efflux liability, but the CSF exposure surrogate component is absent. What would change the answer is a candidate-specific nonclinical CSF concentration readout or a central occupancy measurement demonstrating that the measured Kp,uu translates into pharmacologically relevant unbound brain exposure.

Position A's constructive case rests on two precedent comparisons, both of which have broken or unverified comparable bases under record scrutiny. (1) precedent_pc7b071 (dapansutrile): The visible record (PMID:33257576) describes dapansutrile as 'safe in humans' generally and delivers it via oral chow in mice (peripheral-like exposure). It does not report a specific low-hepatotoxicity clinical profile, and more importantly the precedent is peripherally restricted whereas the candidate is brain-penetrant. This divergence in distribution means dapansutrile's safety profile does not speak directly to the hepatic liability of a CNS-penetrant molecule, so the comparator is broken. (2) precedent_p7bbc38 (MCC950): The cited record (PMID:30381407) discusses MCC950's efficacy in Parkinson models but does not mention furan rings, transaminase elevations, or clinical development halts in its abstract. Position A's claim that the candidate is 'structurally differentiated' from the MCC950 furan-liability is therefore unanchored in the visible record; the record offers no structural rationale showing the candidate escapes the same liability. The measured result that both sides accept—a mild, reversible, dose-dependent transaminase elevation confined to the high-dose group in a repeat-dose rodent study—is encouraging, but on the DMPK dimension the claim requires confidence that the liability is structurally resolved and that an adequate exposure margin exists. Position B correctly identifies that the record lacks both a demonstrated structural basis and a margin between projected efficacious unbound exposure and the hepatic NOAEL. The authored negative (negative_longest_exposure_twelve_weeks) further constrains translational confidence for chronic Alzheimer dosing, though this is a class-level limitation rather than a direct rebuttal of the preclinical signal. Because Position B's objections are anchored to real gaps that break Position A's precedent comparisons, while Position A's optimism relies on unverified structural claims and a mismatched peripheral precedent, the skeptical position is better supported.

The criterion dimension asks whether the class hepatic caution is addressed as a structural question rather than merely a monitoring one. Position A concedes that the record lacks an explicit structural demonstration (e.g., absence of a furan or related alert) in the candidate's chemotype, and instead rests on a functional severity gap backed by class-level structural diversity. The functional severity gap (mild, reversible transaminase elevation) is a monitoring framing, not a structural one. Position B correctly notes that dapansutrile represents what structural de-risking looks like in this class—explicitly low hepatotoxicity contrasted with the MCC950 caution—and the candidate does not share that feature. The authored negative (no CNS-penetrant NLRP3 inhibitor dosed beyond twelve weeks in neurodegeneration) further undermines the reversibility claim for chronic AD use. On the structural dimension, the record does not reach far enough to support the claim that the hepatic caution is structurally addressed.

Position B’s skeptical read is better anchored in the record. Position B identifies a specific structural-liability gap—no evidence that the candidate’s chemotype has been evaluated for or cleared of the features implicated in MCC950’s clinical halt—and Position A’s own assessment concedes this exact point: ‘the record does not yet explicitly confirm that the candidate’s chemotype lacks the structural liability attributed to MCC950.’ Under the principle that an open disqualifying objection is not neutralized by a strong constructive case elsewhere, this conceded gap stands. Position A invokes OLT1177 as a favorable precedent, but that precedent earned its hepatic-safety position by demonstrating low hepatotoxicity explicitly contrasted with MCC950—i.e., by structural differentiation. Because the candidate has not demonstrated the same structural differentiation, the precedent comparison breaks on the safety dimension. The mild, reversible rodent transaminase signal is a verified positive finding, yet on the hepatic-sufficiency dimension—where the class caution is rooted in a structure-driven Phase 2 halt—it does not resolve the underlying structural question. Therefore, Position B more accurately reflects how far the record reaches.

The controlling issue is that the claim dimension asks whether the class hepatic caution is addressed as a structural question rather than a monitoring one. Position A correctly establishes a class-level structural reframing: the record supports that MCC950's hepatotoxicity is linked to its furan-ring structure, not to NLRP3 inhibition per se, and Dapansutrile demonstrates the same mechanism can be hepatologically clean. However, per the principle of separating verified from reported, the inference that the candidate has avoided the offending structural feature is exactly that—an inference from pharmacological phenotype, explicitly acknowledged by Position A as resting on an undisclosed structure. Position B's anchored objection that the candidate's own structural basis is unexamined is correct and remains open. An open disqualifying objection is not neutralized by a strong constructive case elsewhere. For the candidate specifically, what exists in the record is a monitoring argument (mild, dose-dependent, reversible transaminase elevation in rodents), not a structural argument. The class-level structural understanding informs interpretation but does not close the candidate-specific structural question. Position A's comparability call on precedent_p7bbc38 ('yes') overstates the transfer from class-level structural de-risking to candidate-level structural verification. Position B's 'no' on precedent_p7bbc38 accurately captures this gap.

The claim rests on two constructive pillars: (1) mild reversible transaminase elevation in one rodent species supports manageability, and (2) class precedent (dapansutrile safety, MCC950 chemotype-specific liability) frames the signal as non-preclusive. Both pillars are real record items, but they do not reach far enough on the hepatic-sufficiency dimension because Position B identified disqualifying gaps that the record leaves open. First, dapansutrile's favorable hepatic profile was established in a peripheral-predominant exposure context; the candidate is brain-penetrant by design, and the record contains no evidence that hepatic safety under chronic CNS-penetrant exposure matches the peripheral precedent. Second, the MCC950 furan-ring structural attribution is general medicinal chemistry knowledge in the record, but the record does not contain candidate-specific structural evidence showing this chemotype avoids that liability. Position A's assertion that 'the structural question has been engaged' is reported, not verified, for this molecule. Third, no exposure-margin analysis linking projected efficacious unbound concentration to the hepatotoxic dose was retrieved, and no second-species repeat-dose data is in the record. Fourth, the negative precedents are verified: the longest published CNS-penetrant NLRP3 inhibitor exposure in any neurodegenerative population is ~12 weeks (ALS; NCT05981040/Usnoflast), and no NLRP3 inhibitor has been dosed in an AD population. The mild reversible rodent signal is a genuine positive, but on the dimension of whether hepatic safety supports DC nomination, the record offers monitoring data without structural mitigation, undefined therapeutic index, and absent chronic exposure precedent. The disqualifying objection therefore stands.

The record itself states that whether the molecule is developable as an oral CNS agent is 'unresolved at this gate,' which Position B anchors directly. Position A counters with class-level precedents—dapansutrile, VENT-02, NT-0796, and VTX3232—that verify oral CNS delivery of NLRP3 inhibitors is feasible, but none of these precedents address the candidate-specific dependency on an unmanufactured metastable amorphous dispersion with recrystallization risk. The precedents therefore establish a viable mechanistic development path but do not bridge the specific CMC/formulation gap for this candidate. Both sides invoke DC-gate threshold standards that are background knowledge, not record-anchored. Because the record explicitly labels the question unresolved and the precedents do not resolve the candidate-specific formulation liability, the evidence does not resolve the claim in either direction.

Position B carries the stronger anchoring on the CMC-and-supply dimension. It identifies three record-anchored gaps that directly affect GLP toxicology interpretability: (1) the enabling amorphous dispersion is unmanufactured, leaving salt/polymorph and solid-state stability uncharacterized; (2) recrystallization risk threatens exposure reproducibility between efficacy and toxicology studies, which would undermine the margin-of-safety interpretation that GLP tox requires; (3) the readout itself frames oral CNS status as 'unresolved at this gate' and notes a risk of form-limited dose escalation. These are candidate-specific liabilities. Position A offsets them with class precedent (VENT-02, NT-0796, VTX3232, dapansutrile), but the comparability basis for those precedents is matched mechanism and modality, not matched CMC readiness or solid-state stability. The precedents show that oral CNS delivery is achievable for other NLRP3 inhibitors; they do not show that the candidate's unmade amorphous dispersion is sufficiently characterized to commit to a GLP tox study whose margin interpretation depends on predictable exposure. Dapansutrile, the shared precedent, is explicitly peripheral-oriented and lacks oral-CNS formulation data, so it does not close the gap. Position A's argument that these issues are to be managed during IND-enabling work is reasonable for some CMC dimensions, but it does not neutralize the specific objection that the tox study's exposure-response relationship may be compromised by recrystallization before the study begins. On the evidence in the record, the claim of developability sufficiency is more strongly challenged than supported.

Position A is better anchored in the retrieved record. Its three converging lines are each tied to specific, verified publications: (1) SDD technology maturity and crystallization control (PMID:39486520, PMID:37865133, PMID:37792707); (2) class-level demonstration that oral CNS-penetrant NLRP3 inhibitors are pharmacologically achievable (PMID:42062792/PMC:13132802 for VENT-02 CSF exposure; PMID:37874905 and PMID:40792655 for NT-0796 brain penetration and clinical CSF exposure); (3) the readout framing the dispersion as a formulation strategy rather than an intractable barrier. Position A also explicitly scopes its claim carefully, conceding that the enabling dispersion has not been manufactured and that toxicology-scale stability is unproven. This concession does not defeat its claim; it limits it to the proposition that these liabilities do not preclude nomination, which the precedented technology and class-level achievements support. Position B's disqualifying objection—that absent salt/polymorph selection, solid-state stability, and pilot-scale synthesis make GLP commitment impossible—is an inference from absence that Position B does not tie to any specific record item showing such gaps are disqualifying at the Development Candidate gate. Position B also fails to engage Position A's additional precedents (VENT-02, NT-0796, VTX3232), leaving their comparability unchallenged. The strongest open objection remains the unmanufactured dispersion and unproven toxicology-scale stability; this gap is real and acknowledged by Position A, but the record does not support Position B's stronger conclusion that it is an unresolved barrier to GLP execution. The verified literature supports a credible, precedented path to address the liability, which is the standard expected at nomination, not completion of all CMC work.

On the dimension of developability_sufficiency, the record establishes two verified facts that do not align. First, the candidate's developability is explicitly flagged as unresolved because its enabling amorphous dispersion has not been manufactured, creating recrystallization risk and an unbuilt exposure foundation for IND-enabling toxicology. Second, all four precedents represent NLRP3 inhibitors that achieved human CSF exposure and central target engagement—but from resolved formulation states (chow enrichment in mice for dapansutrile; conventional oral formulations with demonstrated human PK for VENT-02, NT-0796, and VTX3232). Position A argues these precedents show that P-gp and borderline CNS physicochemical properties are 'compatible with' achieving CNS exposure, but the retrieved record contains no data on the precedents' P-gp status, CNS-MPO scores, formulation requirements, or solid-state stability. Without those profiles, the precedents cannot serve as direct comparators for the candidate's specific unresolved developability state. The common_scale_basis Position A asserts—'shared developability challenge'—is not present in the retrieved record for any precedent. Position B correctly identifies that the precedents are at a higher maturity stage on the developability dimension and therefore do not bridge the candidate's gap. The strongest open objection—the unmanufactured amorphous dispersion leaving the CMC and exposure foundation unbuilt—is confirmed by the record and is not neutralized by precedents whose own developability journeys are undocumented. The class feasibility argument (NLRP3 inhibition achieves CNS exposure) is real but operates on a different dimension than developability sufficiency at the DC gate.

The claim under arbitration is human dose projection sufficiency. The verified record explicitly lacks a candidate-specific derivation: no predicted human clearance, no allometric or PBPK basis, no projected clinical dose or regimen, and no stated species basis for the carried efficacious exposure. Position A correctly notes that distributed inputs exist (Kp,uu ~0.22, AD-model efficacy, class-level human CNS PK/PD precedents), but these verify mechanism-class translatability and preclinical model validity, not the specific dose projection required by the claim. Class precedents such as VENT-02, NT-0796, and VTX3232 verify that CNS-penetrant NLRP3 inhibitors can achieve human brain exposure and biomarker engagement, yet they are different molecules with distinct human PK; they cannot substitute for candidate-specific clearance or dose prediction. Dapansutrile validates the APP/PS1 model preclinically but carries no published human AD dose projection. The authored negatives—no NLRP3 inhibitor dosed in AD, no positive controlled neurodegenerative efficacy, and a twelve-week exposure ceiling—are properly scoped as statements about the published clinical record; they remove indication-specific and efficacy anchor points rather than merely bounding confidence. Because the record contains the dose-projection derivation only in inferred, distributed form and not as an explicit candidate-specific prediction, Position B’s skeptical reading is more tightly anchored to what the record verifies versus what it reports by analogy.

The dimension under review is human_dose_projection_sufficiency. The record contains explicit, unambiguous package-level readouts: human_dose_projection_status is 'not derived,' predicted_human_clearance is 'not recorded,' allometric_or_pbpk_basis is 'none recorded,' projected_clinical_dose_and_regimen is 'not recorded,' and projected_fractional_target_coverage_over_the_dosing_interval is 'not stated for either compartment.' These are direct record items, not inferences. Position A does not dispute them; instead, it redescribes distributed pharmacology readouts (CNS penetrance inference, central target engagement, AD-model efficacy) as constituting a sufficient basis for the DC gate. This shifts the dimension from dose-projection derivation to general pharmacologic plausibility. On the actual dimension, the record reaches essentially zero. The precedents Dapansutrile (precedent_pc7b071) and VENT-02 (precedent_pcf4b44) are real molecules with relevant mechanism and modality data, but neither provides a derived human dose projection for Alzheimer's disease, and Position A's own comparability language concedes that Dapansutrile 'still provides no human dose projection or three-pillar argument for AD.' The field negatives (no NLRP3 inhibitor dosed in AD, no positive controlled neuro efficacy, longest exposure 12 weeks) are bounded-scope curated statements about the clinical record; within their stated scope they correctly indicate that no clinical anchor exists to calibrate a chronic AD dose projection for this class. Position B's disqualifying objection—that the efficacious exposure's species basis is unstated, rendering the chronic safety margin an unmoored ratio—is anchored in the record and is not rebutted by Position A, which concedes the species basis is unstated while simultaneously calling the margin 'anchored.' That contradiction is a weakness in Position A's anchoring. Because the package is blank on every element of the dimension and the precedents do not bridge that gap for this indication, Position B more accurately characterizes how far the record supports the claim.

Position B's skepticism is more firmly anchored to the record. Both sides agree the package lacks a formally derived human clearance prediction, allometric or PBPK basis, projected clinical dose and regimen, and stated fractional target coverage. Position B additionally identifies an unstated species basis for the carried efficacious exposure, which if mismatched would invalidate the chronic-dosing safety margin rather than merely leave it incomplete. Position A counters with class-level clinical CNS engagement data (VENT-02, NT-0796, VTX3232), which are verified in the record. However, these precedents are anchored in Parkinson's disease or healthy-volunteer populations, not Alzheimer's disease. The cross-indication translation from PD clinical exposure-response to AD dose projection is plausible but not verified in the record; Position A treats it as a 'credible translational path,' yet no record item establishes that PD NLRP3 engagement predicts AD efficacious exposure. The authored negatives (no AD dosing, no positive neurodegenerative efficacy readout, limited exposure duration) are curated statements about the published record and are credited within their stated scope. On the dimension of human dose projection sufficiency specifically, the verified absence of AD-specific clinical PK/PD anchors and the unverified species basis for the efficacious exposure weigh more heavily than the PD-class scaffold Position A provides.