Skip to content

comp-039 — CFH-dependence mechanism-dissociation of dietary upstream-CP0 candidates

Plain-English summary first. Open Enzyme has identified four dietary candidates that block complement activation upstream of MSU-crystal-driven gout flares: rosmarinic acid (in rosemary, lemon balm, perilla), luteolin (in celery, parsley, chamomile), Houttuynia cordata polysaccharide (in fish-mint, a herb commonly eaten across Southeast Asia), and Helicteres angustifolia benzofuran lignans (a tropical folk-medicine plant — not dietary). A common genetic variant called CFH Y402H is present in ~36-39% of Europeans, ~35-37% of Africans, ~30% of South Asians, ~5-6% of East Asians. It weakens one specific complement-regulator protein (Factor H). The OE prediction is that Y402H carriers should benefit more from these dietary candidates because the candidates work upstream of where Factor H normally acts — they prevent the problem before Factor H would have to clean it up. There is one big counter-evidence: in age-related macular degeneration (AMD), Y402H carriers paradoxically did worse on the AREDS zinc-antioxidant formulation. The OE hypothesis is that AREDS works through Factor H (and carriers can't capitalize on it because their Factor H is broken), while these dietary candidates work upstream of Factor H (so carriers can still benefit). This comp converts that hand-wavy plausibility argument into a candidate-by-candidate prediction the UK Biobank cross-tab can test.

Where the analysis lives: - Operations workspace: operations/cfh-mechanism-dissociation-2026-05-21/ - Per-candidate Model A source reads + Model B (DeepSeek) counter-reads + two-model annotated cross-checks: operations/cfh-mechanism-dissociation-2026-05-21/outputs/ - Biobank feasibility (prior): logs/cfh-y402h-dietary-cp0-biobank-mining-2026-05-19.md - Canonical CFH-row source-of-truth: gout-genetic-variants.md Category 5; complement-c5a-gout.md §6.3

1. Scope and question

CFH (Complement Factor H) Y402H — the missense variant rs1061170, p.Tyr402His — sits in Sushi/CCP domain 7 of CFH (UniProt P08603 verified: Sushi 7 = aa 387-444 [grep-verified against the P08603 JSON record this session]; Y402 is in the middle of that domain). CCP6-8 of CFH is the canonical CRP-binding and host-glycosaminoglycan-binding surface. Y402H weakens both surface contacts. CFH itself regulates alternative-pathway complement amplification by (a) decay-accelerating the C3bBb convertase and (b) acting as a Factor-I cofactor for C3b → iC3b cleavage. Both functions operate on surface-deposited C3b — that is, downstream of where C3b is first generated and deposited.

The OE hypothesis (complement-c5a-gout.md §6.3, added 2026-05-19): the comp-018 / comp-020 upstream-CP0 dietary candidates inhibit complement upstream of where CFH acts (preventing C3 convertase assembly or C3b deposition), so Y402H carriers — who have more unregulated surface C3b at baseline — should benefit more from these candidates, not less. The AMD-paradox counter-evidence (Klein 2008 PMID 18423869 / Awh 2013 PMID 23972322 / Vavvas 2018 PMID 29311295 AREDS zinc + Merle 2015 PMID 26132079 DHA → CFH high-risk carriers WORSE) is hypothesized to apply only to interventions that work through CFH (zinc-induced complement inactivation requires CFH-CRP-bridging, which Y402H performs poorly), not to interventions that work upstream of CFH.

This comp converts that hypothesis from hand-wavy plausibility to a per-candidate CFH-dependence classification with falsifiable predictions.

2. Methodology

Per CLAUDE.md §"Multi-frame query discipline" + §"Pre-commit grep-verify gate for load-bearing numbers" + §"Translation protocol":

  1. Multi-frame retrieval (PubMed + East Asian academic sources via local-curl allowlist; query plans at operations/.../inputs/per-candidate-query-plan.json; reachability probes at operations/.../outputs/retrieval-probes.json).

  2. Two-model independent cross-check. Model A = Claude Opus 4.7 (this comp's primary reader, full primary-source reads); Model B = DeepSeek (deepseek/deepseek-chat-v3 via OpenRouter, native-Chinese training depth, independent counterread per candidate). Per CLAUDE.md update 2026-05-21: Claude is Model A and reads sources directly; OpenRouter is invoked only for Model B to control marginal cost.

  3. Binding-site mapping vs Y402H footprint. For each candidate, the primary-source binding-site or target-identification data is mapped onto the complement cascade, and the question "does this mechanism require functional CFH downstream?" is answered with structural justification.

  4. Multi-hypothesis discipline (per CLAUDE.md §"Deep multi-metric evaluation discipline"). For each candidate, ≥2 binding-site / mechanism hypotheses are generated, evaluated against published evidence, and the rejected ones are documented.

  5. Pre-commit grep-verify gate. All load-bearing numbers (CH50, AP50, IC50, residue ranges, sample sizes, depletion-rescue percentages) verified against the primary source before writing. Sushi-7 = aa 387-444 verified against UniProt P08603 JSON directly. The Lu 2018 depletion-rescue percentages (CHCP C3-depleted 9.29 ± 1.69%, C4-depleted 12.34 ± 1.39%, C5-depleted 44.54 ± 3.92%) verified against PMC5925397 full text. The Yin 2016 CH50 / AP50 (compound 4 = 40 / 105 μM; compound 5 = 9 / 21 μM) verified against PMC6273495 full text. The Sahu 1999 IC50 = 34 μM verified against PubMed-hosted abstract.

3. Per-candidate classification

3.1 Rosmarinic acid

Classification: CFH-INDEPENDENT (High confidence — both models agree).

Primary mechanism evidence: - Sahu 1999 (PMID 10353266): rosmarinic acid covalently attaches to the thioester-containing α'-chain of nascent C3b (Cys988 reactive carbonyl, exposed only on activated/cleaved C3). IC50 for inhibiting covalent C3b deposition on cells = 34 μM. Mechanism is fluid-phase covalent capture of nascent C3b before surface attachment. [In Vitro] - Englberger 1988 (PMID 3198307): inhibits C3-convertase of the classical complement pathway. Threshold 1 μM; optimal 5-10 μM, ~70% hemolysis inhibition. Mechanism is convertase active-site / assembly inhibition (elastase weakly co-inhibited supports serine-proteinase active-site involvement). [In Vitro] - Peake 1991 (PMID 1761351): mM-range C5 convertase inhibition; secondary regime. [In Vitro]

CFH-footprint mapping: the C3b thioester (Cys988 α'-chain) is structurally distant from C3d (the CFH-binding fragment of C3b) and far from CFH's CCP6-8 CRP-binding surface. Rosmarinic acid quenches nascent C3b before it can be deposited on a target surface; CFH's regulatory functions (decay-acceleration of C3bBb, Factor-I cofactor for C3b → iC3b) operate on already-surface-deposited C3b — downstream of where rosmarinic acid acts. No primary source mentions CFH dependence.

Predicted Y402H × rosmarinic acid × incident gout interaction: negative direction (effect ≥ in carriers) — Model A reasoning. Model B picks null direction (no differential effect). Both reject the AMD-paradox direction (+ harm in carriers).

[TRANSLATION-DISAGREEMENT — prediction direction]: Model A predicts effect ≥ in carriers because Y402H carriers have more unregulated surface C3b at baseline; removing the upstream substrate produces a larger absolute reduction. Model B predicts null because mechanism independence implies genotype indifference. Both predictions need separate UKB falsification thresholds; positive direction (carriers worse) refutes both.

Multi-hypothesis discipline — rejected alternatives: 1. Rosmarinic acid acts via CRP displacement at the CFH CCP7 surface. Rejected: Sahu 1999 IC50 was measured in cell-free fluid-phase nascent C3b — no CRP, no host surface, no CFH present; mechanism works in their absence. 2. Rosmarinic acid binds C3d (the CFH-binding fragment) directly. Rejected: radioiodination localizes covalent attachment to the α'-chain thioester (Cys988), not C3d (the TED domain).

Falsification test: rosmarinic acid (34 μM, 100 μM, 340 μM) on MSU-crystal-driven complement activation in CFH-depleted serum vs CFH-replete serum (C5a generation readout). Retained suppression in CFH-depleted serum confirms CFH-independence.

UKB cross-tab specification: rs1061170 × Phenol-Explorer-derived rosmarinic-acid intake (rosemary, perilla, lemon balm sources via Oxford WebQ — note these are Mediterranean-pattern foods and exposure is heterogeneous) × incident gout M10.x. Pre-specify null AND negative direction with adequate power for each.

3.2 Luteolin

Classification: CFH-INDEPENDENT (Medium confidence — both models agree, both flag mechanism-site ambiguity).

Primary mechanism evidence: - Zhang 2008 (PMID 18400428 / PMC7126446, full text read): luteolin tested in matched CP (sheep-erythrocyte + guinea-pig serum) and AP (rabbit-erythrocyte + 1:10 NHS) hemolytic assays. CH50 = 0.19 mM; AP50 = 0.17 mM. [Verbatim verified against PMC7126446.] Mechanism attribution: 4'-OH B-ring substitution essential for activity (flavonoids with -OCH3 at C-4' are 5-10× weaker). Specific cascade node NOT pinned at depletion-rescue resolution. [In Vitro] - Pieroni 1996 (PMID 8941947): olive-leaf flavonoid class-level activity. [In Vitro]

CFH-footprint mapping: the matched CP/AP IC50 (0.19 vs 0.17 mM) is consistent with a cascade node common to both pathways — most likely C3 itself or convertase active-site / assembly. Both candidate mechanisms are functionally upstream of CFH's AP-specific regulatory role. A CFH-competitive mechanism would predict AP-selective potency (CFH is AP-specific), which is not observed.

Luteolin additionally has documented gout-relevant non-complement modes (per comp-013): XO IC50 550 nM + URAT1 expression downregulation. These are CFH-orthogonal entirely. The multi-mode complicates the UKB cross-tab — see "Limitations" below.

Predicted Y402H × luteolin × incident gout interaction: negative direction (effect ≥ in carriers, Model A) vs null direction (Model B). Both reject the AMD-paradox direction.

[TRANSLATION-DISAGREEMENT — mechanism specificity]: Model A commits to "convertase-level inhibition" as the most-likely mechanism site; Model B reserves judgement. Model B's more cautious framing is appropriate given that no depletion-rescue or single-residue binding data exists for luteolin specifically.

Multi-hypothesis discipline — rejected alternative: 1. Luteolin binds the CCP6-8 CRP-binding surface of CFH itself, competing with CRP. Weakly rejected: (a) no surveyed paper reports CFH as a luteolin target; (b) the matched CP/AP IC50 is inconsistent with a CFH-competitive mechanism (CFH-competition predicts AP-selectivity, not matched CP+AP potency).

Falsification test: luteolin on MSU-crystal-driven complement activation in CFH-depleted vs CFH-replete serum. For the UKB cross-tab, additionally pre-specify a 24h-urate intermediate-phenotype arm to dissociate the complement-mode from the XO + URAT1 urate-axis modes — otherwise the cross-tab is confounded by three parallel mechanisms.

UKB cross-tab specification: rs1061170 × Apiaceae-family dietary intake (celery, parsley, chamomile — captures luteolin exposure better than total-flavonoid sums) × incident gout M10.x, with 24h-urate intermediate readout in a subset.

3.3 Houttuynia cordata polysaccharide (HCP / HCPM / CHCP)

Classification: CFH-INDEPENDENT (High confidence — both models agree).

Primary mechanism evidence: - Lu 2018 (PMID 29719782 / PMC5925397, full text read): complement-depleted-sera mapping for CHCP polysaccharide. Verbatim depletion-rescue percentages: C2-depleted serum rescue 72.82 ± 10.61%, C9-depleted rescue 63.21 ± 2.27%, C3-depleted rescue 9.29 ± 1.69%, C4-depleted rescue 12.34 ± 1.39%, C5-depleted rescue 44.54 ± 3.92%. Interpretation: CHCP's primary targets are C3 and C4, with partial C5 effect. [Grep-verified against PMC5925397.] [In Vitro + Animal Model — rats, "two-hit" ALI rescue] - Tian 2014 (PMID 24423008): glycoside-class anti-complementary fractions of Houttuynia cordata "block C3 and C4 components" — convergent mechanism class. [In Vitro] - Xu 2015 (PMID 26190353 / PMC7127486, full text read): HCP at 40, 80, 160 mg/kg PO in LPS-induced ALI mouse model — reduces C3d deposition (immunohistochemistry), reduces TLR4 expression in lung tissue, suppresses C5a-induced macrophage chemotaxis, reduces NLRP3-axis cytokines (IL-1β, TNF-α, IL-6). Multi-target: complement + TLR4 + NLRP3 axis. [Animal Model] - Yu 2026 PMC12937656 (TLR4-MD2 computational docking) — CP1-level TLR-priming mechanism orthogonal to complement.

CFH-footprint mapping: C3 binding upstream of cleavage prevents the substrate pool that CFH regulates. C4 binding is mechanistically incompatible with CFH-dependence — CFH is alternative-pathway-specific and does not regulate the classical-pathway C4-axis. The TLR4-MD2 mode (CP1, TLR-priming) bypasses complement entirely. Pectic polysaccharide binding mode (adsorptive surface contact) is structurally distinct from CFH's CCP-fold protein recognition. Three orthogonal lines of evidence converge on CFH-independence.

Predicted Y402H × HCP × incident gout interaction: negative direction, possibly notably greater in carriers (Model A — dual CP0+CP1 chokepoint compounds the genotype-baseline-severity amplification) vs null direction (Model B). Both reject the AMD-paradox direction.

[TRANSLATION-DISAGREEMENT — dual-chokepoint framing]: Model A explicitly highlights HCP's dual CP0 (complement) + CP1 (TLR4) mechanism as a reason for amplified benefit in Y402H carriers; Model B does not foreground this. The dual-chokepoint framing is consistent with the comp-018 Phase 2 "DUAL-CHOKEPOINT" classification of HCP/HCPM as a Tier 1d candidate.

Multi-hypothesis discipline — rejected alternative: 1. HCP polysaccharide acts as a CFH-mimic — binding C3b at the same surface CFH binds. Partially rejected: depletion-rescue data (Lu 2018) shows HCP blocks the cascade at C3 cleavage, not at C3b inactivation. CFH-mimicry would predict downstream-of-cleavage action. HCP's pectic polysaccharide structure is structurally nothing like CFH's CCP-fold. Same prediction direction (negative) but better-evidenced as activation-blocker than CFH-mimic.

Falsification test: CHCP / HCP-1 on MSU-crystal-driven complement activation in (i) CFH-depleted serum, (ii) Factor B-depleted serum, (iii) C4-depleted serum. The C4-depletion arm specifically tests the classical-pathway-CFH-independent prediction.

Cross-tab specification — NOT well-suited to UK Biobank. Houttuynia consumption is rare in UK dietary corpus. The relevant cross-tab is more tractable in Chinese / Vietnamese / Korean / Japanese cohorts — Korean Genome Epidemiology Study (KoGES), China Kadoorie Biobank (CKB), Singapore Chinese Health Study. These cohorts have lower Y402H allele frequency (~5-6% East Asian vs ~36-39% European) but adequate Houttuynia exposure data. Operational recommendation: if UKB is the active collaboration channel, defer HCP cross-tab and lead with rosmarinic acid + luteolin; if a parallel East Asian cohort collaboration becomes available, HCP becomes the highest-priority cross-tab there.

3.4 Helicteres benzofuran lignans

Classification: CFH-INDEPENDENT (Medium confidence — Model A bounds by replication risk; Model B picks High because Yin 2016 target identification is unambiguous if it replicates).

Primary mechanism evidence: - Yin 2016 (PMID 27834928 / PMC6273495, full text read): depletion-rescue target identification for Helicteres angustifolia benzofuran sesquilignans. Compound 4 (machicendonal): CH50 = 40 μM, AP50 = 105 μM; targets C1q, C2, C3, C4, C9. Compound 5 (dihydrodehydrodiconiferyl alcohol): CH50 = 9 μM, AP50 = 21 μM; targets C1q, C2, C3, C9 (NOT C4). [Verbatim verified against PMC6273495.] [In Vitro, single-paper anchor] - comp-018 Phase 2 replication status: INCONCLUSIVE (phase-2-helicteres-replication.json) — no independent group has reproduced Yin 2016. Structurally-adjacent benzofuran lignans (Styrax japonica egonol, Min 2004 PMID 15643559) are 3.7× weaker, leaving open the question of whether Yin 2016 represents exceptional pharmacology or assay-format artifact.

CFH-footprint mapping: C1q, C2, C4, C9 are not CFH-regulated. C3 binding is upstream of cleavage (CFH regulates the cleavage product, not C3 itself). The multi-target lignan pattern is structurally orthogonal to CFH's CCP6-8 binding surface. The CP-pathway IC50 (9 μM CH50 for compound 5) is more potent than the AP-pathway IC50 (21 μM) — inconsistent with a CFH-competitive mechanism (CFH is AP-specific; competition would predict AP-selectivity).

Predicted Y402H × Helicteres × incident gout interaction: negative direction (effect ≥ in carriers, Model A) vs null (Model B), with wide uncertainty bands. Practically: Helicteres is not part of any UKB-tractable dietary corpus. The candidate's CFH-classification value is more relevant to a future TCM-supplement clinical evaluation than to the present UKB cross-tab.

[TRANSLATION-DISAGREEMENT — confidence framing]: Model A bounds confidence by comp-018 Phase 2 replication risk (Medium); Model B does not weight replication risk into the confidence call (High). Model A's framing is the appropriate one for OE's evidence discipline — replication risk is a published Phase 2 finding, not informal speculation.

Multi-hypothesis discipline — rejected alternative: 1. Helicteres lignans act via CCP6-8 occupation displacing CRP / GAG from CFH. Rejected: Yin 2016 depletion-rescue identifies C1q + C2 + C3 + C4 + C9 (not CFH) as targets; CP-pathway IC50 is more potent than AP, inconsistent with CFH-competitive mechanism.

Falsification test priority order: 1. Independent wet-lab replication of Yin 2016 (comp-018 Phase 2 open follow-up). Load-bearing. 2. If replication confirms, compound 5 on MSU + CFH-depleted vs CFH-replete serum. 3. Consider whether Helicteres has a route to clinical translation given non-dietary status.

4. Summary table — per-candidate genotype × candidate interaction prediction

Candidate CFH-dependence Confidence Predicted Y402H × candidate × incident gout (Model A) Predicted (Model B) Falsification threshold
Rosmarinic acid CFH-independent High negative (effect ≥ in carriers) null HR > 1.5 (402HH × diet vs 402YY × diet) at p<0.05 refutes — retire from CP0 stack
Luteolin CFH-independent Medium negative (effect ≥ in carriers, with multi-mode confound) null HR > 1.4 refutes; cross-tab against Apiaceae intake + 24h-urate intermediate readout
Houttuynia cordata polysaccharide (HCP/HCPM/CHCP) CFH-independent High negative, possibly notably greater (dual CP0+CP1 chokepoint) null HR > 1.3 refutes; UKB unlikely to capture exposure — defer to East Asian cohorts
Helicteres benzofuran lignans CFH-independent Medium (replication-bounded) negative, wide uncertainty null (a) wet-lab replication of Yin 2016 first; (b) non-dietary, not UKB-actionable

Mechanism axis common to all four candidates: action at C3 / C3 convertase / C3b nascent thioester / C1q + C2 + C4 (CP initiation) — all upstream of where Y402H matters. Mechanism axis present in zero of the four candidates: zinc-induced complement inactivation via CFH-CRP-bridging (the AREDS / DHA AMD-paradox mechanism). The dissociation is mechanistically clean: OE candidates work by substrate-elimination upstream of CFH regulation, AREDS works through CFH-CRP-bridging-dependent complement inactivation. Y402H breaks the latter but not the former.

Updated 2026-05-22 (sweep 2026-05-21 Contradiction #1 + Experiment #3). The original specification led with UK Biobank as the primary cross-tab cohort. That framing under-powers the cross-ancestry direction-check: CFH Y402H allele frequency is comparable in European (~36–39%) and African (~35–37%) populations (per gout-genetic-variants.md Category 5), but the UK Biobank cohort is European-ancestry-skewed, and Volcik 2008 (ARIC) found the CFH 402H × hypertension interaction was significant in whites but null in African-Americans — i.e., the variant has population-specific effect-direction differences for a related outcome. A UKB-only cross-tab would be well-powered for the European signal but provide no cross-ancestry direction-check for the African-ancestry population where the variant is equally common and the effect-direction may differ. The corrected framing: UK Biobank and All of Us as co-primary cohorts, not UKB as primary with AoU as a parallel side-check. The All of Us Researcher Workbench has substantially better African-American representation, lower credentialing friction (~2–4 weeks, free), and is the appropriate cross-ancestry direction-check counterpart to UKB. Two independent cross-tabs in two cohorts with comparable CFH 402H frequency but different ancestry composition and dietary patterns produce the most informative answer.

For the UK Biobank gout-GWAS collaborators (Merriman/Otago, Major-Wrigley/Auckland, Choi/MGH per gout-genetic-variants.md Category 5 CFH row) AND for the All of Us Researcher Workbench cross-tab, the candidate-stratified specification is the following:

Lead query (highest-power tractable cross-tab):

Cross-tabulate rs1061170 (CFH Y402H) genotype × Phenol-Explorer-derived rosmarinic-acid intake quartiles × incident gout M10.x in UK Biobank participants without prior gout at baseline (n target ≈ 450K). Pre-specify both a negative-direction test (carrier-amplified protective effect, HR < 1 in carriers vs non-carriers on high-intake diet; Model A) AND a null-direction test (no differential interaction; Model B). The AMD-paradox direction (carriers WORSE on high-intake, HR > 1.4) refutes the OE upstream-CP0-bypass hypothesis and would force retiring rosmarinic acid from the upstream-CP0 candidate stack.

Secondary queries (in priority order):

  1. rs1061170 × Apiaceae-family intake × incident gout M10.x (luteolin proxy; Apiaceae captures luteolin-rich exposure better than total-flavonoid sums). Add 24h-urate intermediate-phenotype readout in the subset with biochemistry data to dissociate the complement-mode from the XO + URAT1 urate-axis modes.

  2. rs1061170 × dietary CFH-bypass diversity score × incident gout (Bondonno 2025 Nature Food methodology generalized) — composite score over rosemary / lemon balm / perilla / celery / parsley intake, vs Yokose 2024 Rheumatology cohort framework.

  3. NOT in UKB: Houttuynia cordata cross-tab. The exposure is rare in UK dietary corpus. Defer to Korean Genome Epidemiology Study (KoGES) / China Kadoorie Biobank (CKB) / Singapore Chinese Health Study collaborations when available — flag that the East Asian cohorts have lower Y402H allele frequency (~5-6%) but adequate Houttuynia exposure data.

  4. Not actionable in any biobank yet: Helicteres benzofuran lignans (non-dietary; comp-018 Phase 2 replication required first).

Co-primary cohort allocation (UKB ↔ AoU):

  • UK Biobank — primary for the European-ancestry signal. Lead query (rosmarinic acid × incident gout) executes first there because UKB's Phenol-Explorer-derived rosmarinic-acid intake quantification is the most mature; existing gout-GWAS collaborators have the extraction pipeline; n target ≈ 450K provides the highest absolute power for the dominant-ancestry result.
  • All of Us — co-primary for the African-American ancestry direction-check. Same lead query, replicated in AoU specifically to test whether the European-ancestry direction holds, inverts, or is null in African-American carriers. Volcik 2008's effect-direction heterogeneity for the related CFH 402H × hypertension outcome means this is not an optional confirmation pass — it's a load-bearing falsification gate for the upstream-CP0-bypass hypothesis as a cross-ancestry claim. AoU credentialing is ~2–4 weeks, free, and the Researcher Workbench provides direct query access without needing a wet-lab collaborator's extraction pipeline.
  • Cross-cohort integration rule: report both results separately first (no pooled meta-analysis until both have landed individually), then run an ancestry-stratified meta only if the direction agrees. If the direction differs (e.g., negative in European, null or positive in African-American), the cross-tab terminates at the per-ancestry result and the hypothesis is revised — not averaged. Negative direction in only one ancestry is a positive finding for that ancestry's clinical application, not a refutation of the whole hypothesis; null in only one ancestry is a refinement, not a refutation.

Confound considerations for the collaborators: (a) CFH 402H × hypertension interaction (Volcik 2008 ARIC) and CFH 402HH × elevated CRP (Hecker 2023) mean the cross-tab needs careful adjustment for BMI, baseline CRP, eGFR, hypertension; (b) overall dietary-quality / Mediterranean-diet-pattern correlation with rosmarinic-acid-rich foods needs adjustment; © effect-direction differences in African-American vs European-ancestry (Volcik 2008) are the rationale for the AoU co-primary design — stratifying within a single cohort is inferior because the cohort's ancestry mix introduces residual confounding even after PC-based stratification.

6. Limitations

  1. No direct empirical Y402H × candidate × gout data exists in the published literature. All four classifications are mechanistic extrapolations from primary-source binding-site or target-identification data — the empirical test is exactly what the UKB cross-tab is designed to provide.

  2. AMD-vs-gout site-of-action confounding. AMD complement activation is at the retinal pigment epithelium; gout complement activation is at MSU crystals in joints. The CFH-CRP-zinc bridging machinery may not be identically deployed at both sites, so the AMD-paradox-doesn't-transfer reasoning carries some site-specificity uncertainty.

  3. Bioavailability of dietary candidates is variably characterized. Rosmarinic acid reported low systemic absorption (≤1% per PMC9143754); luteolin variable (Apiaceae sources better than supplementation); HCP polysaccharide bioavailability after oral intake incompletely characterized; Helicteres benzofuran lignans not relevant (non-dietary). The operative in-vivo concentration at MSU-crystal sites in joints is unknown for all four. Gut-luminal or systemic-plasma routes may differ.

  4. Single-anchor replication risk for Helicteres. comp-018 Phase 2 verdict on Yin 2016 is INCONCLUSIVE. The Helicteres CFH-independence classification is conditional on the Yin 2016 target identification holding up in independent replication.

  5. Mechanism-site resolution varies across candidates. Rosmarinic acid (high, Sahu 1999 radioiodination); HCP / Helicteres (medium, depletion-rescue resolution per Lu 2018 + Yin 2016); luteolin (low, no depletion-rescue or single-residue data). Confidence calls reflect this gradient.

  6. Two-model agreement on classification but disagreement on prediction direction. Both models classify all four candidates as CFH-independent. Both reject the AMD-paradox direction (carriers WORSE). They disagree on whether to predict negative direction (Model A — carriers benefit MORE because genotype-baseline severity amplifies absolute effect size) or null direction (Model B — mechanism independence implies genotype indifference). The UKB cross-tab should be pre-specified to falsify both. A positive (carriers worse) result refutes both models and would force retiring the upstream-CP0-bypass hypothesis.

  7. Population-stratification confound for Y402H frequency. Y402H is ~36-39% in Europeans, ~35-37% in Africans, ~30% in South Asians, ~5-6% in East Asians. The UKB cohort is European-ancestry-skewed. Replication across ancestries needs separate analyses, ideally collaborations with All of Us (better African-American representation) and the East Asian cohorts above for HCP.

7. New follow-ups surfaced

  • comp-040 (proposed): wet-lab in-vitro CFH-depleted-serum MSU-crystal complement-activation assay — definitive falsification test of the CFH-independence classification for rosmarinic acid, luteolin, and HCP. Estimated cost: low (assay reagents commercially available; serum sourcing via CompTech or Complement Technology Inc).
  • comp-041 (proposed): Y402H × candidate cross-tab feasibility scan in East Asian cohorts (KoGES, CKB, Singapore Chinese Health Study) for Houttuynia-specific cross-tab — paralleling the UKB feasibility analysis done 2026-05-19.
  • Open follow-up — comp-018 Phase 2 Helicteres replication: unchanged; this comp does not displace the load-bearing replication requirement on Yin 2016.

8. Cross-references