Computational Experiments¶
Tracking index for computational analyses in the Open Enzyme platform. Distinct from validation-experiments.md (wet-lab), these use structure prediction, sequence analysis, and simulation to generate evidence-based priors before committing wet-lab resources.
Convention: Each active or bounded analysis lives at wiki/etc/experiments/comp-NNN-<slug>/ with the code/input/output contract required by its current verdict. A fully invalidated COMP may retain only a hash-bound invalidation record in the live tree while Git preserves the retired executable artifacts.
Peer review: Runnable COMPs declare their reproduction command and dependencies. Invalidated tombstones are not rerun; their historical artifacts are available through Git. Disagreements should be filed as GitHub issues against the relevant comp-NNN folder.
Relationship to wet-lab experiments: Computational analyses inform priors; they shift confidence before a wet-lab experiment runs, and help interpret results after. They do not replace wet-lab validation.
Analyses¶
comp-050 — Luminal UOX Conditional Capacity + Measurement Identifiability (2026-07-28)¶
Question: Which UOX capacity statements are conditionally valid, and which measurement combinations can structurally identify local UOX removal, close the local urate ledger, or attribute removal to systemic-origin urate?
Verdict: METHOD_MAP_DERIVED; biological regime NOT_EVALUATED. The exact rational model derives an inflow-normalized conditional capacity identity and demonstrates that the same luminal-urate concentration trajectory can conceal different UOX-removal fluxes. It does not estimate a human regime, dose, production target, serum effect, topology, chassis, efficacy, assay performance, or safety.
Measurement result: Initial/terminal urate plus a fully qualified product-equivalent observation conditionally identifies local UOX removal but does not close the ledger. Calibrated reaction-site capacity is a separate observation. Both source influxes, boundary fates, and qualified source-resolved product fate make every term in the declared ledger structurally reconstructible, including the unattributed residual; practical closure additionally requires that residual to pass a prespecified tolerance. When product prerequisites fail, local UOX removal and ledger closure become non-identifiable.
Informs: validation §1.33 · H08
Detail: interpretive · experiment folder
comp-049 — Fixed Mixed-Source TCM Evidence Qualification — GATE 1 ONLY / NOT RUN¶
Question: Can four named primary animal studies and one secondary systematic review be qualified into exact-material, polarity-preserving evidence records without reproducing the invalid COMP-013 ranking model?
Status: Gate 1 GO on the exact pre-run snapshot; no result-bearing execution has occurred. The fixed source set and approved design define a bounded replacement experiment. They supply no compound rank, viability verdict, exposure conclusion, target coverage, or advancement decision.
Relationship to COMP-013: COMP-049 does not revive or continue COMP-013’s outputs. It is a new experiment that may re-examine a bounded source set under current evidence and provenance rules.
Next gate: Run only the manifest-bound design, then obtain the mandatory context-isolated post-run review before interpreting or propagating any output.
Detail: pre-run design, code, inputs, and exact review
comp-046 — Dietary Fate Ledger + Endogenous Capture-Fraction Comparison — YELLOW / TWO CONDITIONAL HYPOTHESES (2026-07-13)¶
Question: Two independent comparisons: when does whole-cell GR-5 reduce modeled absorbed dietary precursor, and when does spatial UOX→PDB access exceed an overlap-adjusted well-mixed endogenous capture architecture?
Verdict: YELLOW — two conditional hypotheses, not one additive efficacy claim. The dietary 100-unit fate ledger is conserved. The endogenous side is an architecture-level capture-fraction comparison, not a second conserved fate ledger. Neither is summed into ΔSUA.
Key findings: two independent 81-cell deterministic full-factorials; explicit conserved dietary fate ledger; separate non-conserved endogenous luminal-urate capture-fraction comparison. Grid occupancy is not probability, and the model does not establish a topology winner or joint three-stage efficacy.
Informs: validation §1.34 · purine-degrading bacteria · purine load
Detail: interpretive · experiment folder · Complete first pass
comp-045 — Uricase Topology × Oxygen × Peroxide Design — DESIGN ONLY / NOT EVALUATED¶
Question: How should intracellular+YgfU, LamB-secreted, InakN-displayed, and koji-secreted UOX be compared across urate, oxygen, catalase localization, and VHb support?
Disposition: CANDIDATE_LAYOUT_GENERATED; biological verdict NOT_EVALUATED; wet-lab readiness blocked. Gao/PULSE supplies exact whole-configuration precedents for three EcN topologies with and without the joint KatG+VHb module. It does not isolate KatG or VHb effects, establish extracellular peroxide closure, or supply direct A. oryzae UOX evidence.
Key output: 18 candidate configuration classes, 20 block assignments, 16 preregistered same-block contrasts, and 12 complete 96-well plate maps across three provisional run slots and two planned oxygen contexts whose actual dissolved-oxygen targets must later be predeclared and measured. Every active-UOX well has a support-module-matched inactive-UOX control at the same concentration. Exact constructs, retained activities, oxygen targets, sampling, assay sensitivity, statistical model, decision thresholds, and powered or precision-based run count remain qualification blockers. No topology is ranked or biologically evaluated.
Informs: validation §1.33 · gut-lumen sink · engineered koji protocol
Detail: interpretive · experiment folder
comp-044 — Gut-Lumen Uricase Physiological-Regime Robustness Audit (2026-07-13)¶
Question: Is comp-019's unconditional flat-dose classification robust to explicit substrate occupancy and a finite active window under the inherited priors?
Verdict: COMP-019's unconditional flat-dose classification is not robust to COMP-044's tested substrate-occupancy and finite-window diagnostics. Using the inherited central priors, the 0.59 µM, Km 25 µM, three-hour diagnostic gives capacity ratios 0.093 / 0.466 / 0.932 at 5 / 25 / 50 mg before additional oxygen, access, or survival penalties, versus the legacy 32.3 / 161.7 / 323.4 saturated-capacity calculation. This is an internal-consistency counterexample. It supplies no replacement ΔSUA, dose, genotype order, physiological regime, efficacy model, topology/chassis selection, production-sufficiency target, or safety conclusion.
Key findings: 1,620-cell discrete full-factorial per dose; grid occupancy is not probability; no serum-urate mapping. The 8.3 U/mg activity, Km range, 2–4-hour window, and 233 mg/day denominator are inherited or derived, non-planning-grade inputs. Oxygen, access, survival, and pH attenuation are nonmechanistic scenario multipliers; oxygen stoichiometry and peroxide safety are not modeled. Only the ratio-one boundary has direct meaning within the diagnostic.
Informs: validation §1.33 · H08 · comp-019 interpretation
Detail: interpretive · experiment folder · Complete first pass
comp-043 — EcN Periplasmic Disulfide-Folding Arithmetic — INVALIDATED MODEL (2026-07-13)¶
Question: Which exact-configuration measurements are needed before assigning an EcN configuration to C1-INH, DAF SCR1-4, or lactoferrin?
Verdict: Invalidated and retired. The arbitrary weighting formula and pLDDT-as-accessibility axis cannot establish periplasmic folding capacity, protease survival, viability, priority, or a chassis winner.
Key findings: - The live corpus does not preserve the invalid numerical outputs; Git is the audit trail. - DAF and lactoferrin have distinct disulfide-containing native folds, but annotation does not predict native-fold attainment. Reverify exact feature counts against the current primary record before using them as design inputs. - Single biggest unresolved question: exact-configuration expression, native-fold attainment, secretion, stability, and retained function. No calibrated DsbA/DsbC capacity rule was identified for the proposed secreted configurations. - Compare baseline and DsbC-co-expression arms; either result applies only to the tested construct × route × folding-support configuration.
Informs: engineered-lbp-chassis.md (measurement priorities, not chassis assignment) · chaperone-orthogonal-stacking.md (capacity-calibration gap) · validation-experiments.md
Detail: interpretive · invalidated, non-runnable tombstone · Next gate: exact-configuration expression, native-fold attainment, and retained-function measurement
comp-042 — KPV entry through GSDMD pores vs. a PepT1 baseline — YELLOW, A2 unresolved¶
Question: Does a passive pore model support rapid KPV entry, and what measurements are required to distinguish that route from PepT1-mediated uptake?
Result: A1 is GREEN for intra-articular, YELLOW for subcutaneous, and RED for oral under the declared route-concentration design spaces. These states compare the modeled passive pore contribution with a 10 nM extracellular cell-assay proxy; they do not establish intracellular target engagement or efficacy.
The full A2 sensitivity contains favorable ≥3× heuristic corners: intra-articular crosses in 2/9 moderate-PepT1 cases and 1/9 high-PepT1 cases, while absent/low scenarios cross in all nine cases for every route. A2 nevertheless remains unresolved because the healthy-cell equation, PepT1 scenarios, and matched synovial-macrophage baseline are unvalidated. Concurrent PepT1 transport in the pyroptotic cell is not modeled.
At central pore parameters, the equilibration time constant is 2.17 seconds. At 10 pores and the shortest 60-second lifetime, the modeled fraction is 0.749, so the result does not support “complete equilibration in every ≥10-pore case” or a universal claim that lifetime is irrelevant.
Next gate: validation §1.32 uses an empirically confirmed transporter-orphan tracer and a matched pore-on/off × PepT1-on/off KPV comparator. The wider transporter-orphan pore-delivery hypothesis remains open.
Detail: interpretive page · code, inputs, outputs, and reviews
Former comp-039 — reclassified literature synthesis; no computational verdict (2026-05-21)¶
Question: For rosmarinic acid, luteolin, exact Houttuynia cordata polysaccharide materials, and Helicteres benzofuran lignans, does the candidate's anti-complement mechanism require functional CFH? Candidate identity comes from current evidence homes, not a COMP-018 ranking.
Disposition: This was literature synthesis, not a computational experiment. Its categorical CFH-independence and Y402H-response conclusions are non-authoritative. The cited records still support four candidate-specific Research Conjectures and a matched direct test.
Key findings: - Rosmarinic acid has C3b-directed biochemical evidence; luteolin has whole-pathway CP/AP hemolysis evidence; exact Houttuynia and Helicteres materials have depletion-rescue evidence at multiple complement nodes. [In Vitro] - Those records motivate, but do not prove, retained activity when CFH is absent or impaired. [Mechanistic Extrapolation] - Similar CP/AP potency does not localize luteolin's target. Related Houttuynia fractions cannot be treated as interchangeable. The Helicteres result requires independent matched-material replication. - A diet-by-genotype association can update the tested exposure-proxy hypothesis, not confirm or retire an exact biochemical mechanism. - Next gate: exact-material CFH-depleted/replete/restored serum with MSU activation, proximal C3 and C5a readouts, and assay-interference controls.
Informs: gout-genetic-variants.md Category 5 CFH row · complement-c5a-gout.md §6.3 · upstream-complement-verification-rerun-computational.md
Detail: interpretive
comp-038 — Tier 2 Butyrate Assay Audit — YELLOW (2026-05-20)¶
Question: Is there a Tier 2 butyrate quantification assay (colorimetric, enzymatic, breath-proxy, electrochemical, or other low-cost intermediate method) that can be validated against Tier 3 GC-MS for stool, serum, breath, or culture-supernatant matrices?
Verdict: YELLOW. No ready-to-adopt Tier 1 or Tier 2 butyrate assay was established for current OE use. The scan surfaced a Tier 3 HPLC-UV method for culture supernatants and a separate electrochemical/ANN Tier 2 candidate for stool. Neither has been validated for an OE matrix or workflow.
Key findings:
- PubMed snapshot: 27 queries / 74 records; source snapshot committed at outputs/pubmed-snapshot.json.
- HPLC-UV is a Tier 3 bench method under the current ladder. The De Baere primary abstract (PMID 23542733) supports bacterial-culture-supernatant use, 210 nm detection after ether back-extraction and acidification below pH 2, and matrix-matched calibration from 0.5–50 mM. Transfer into one exact OE strain–medium matrix with spike/recovery and paired GC-MS remains open (validation §1.31).
- Gu 2026 (PMID 42041444) supports a separate electrochemical/ANN stool-specific Tier 2 candidate. The exact published hardware–chemistry–model stack produced butyrate MAE/RMSE/R² of 0.029 mM/0.034 mM/0.998 in a within-study independent 30-sample fecal test cohort versus GC-MS, with a small statistically nonzero negative bias (−0.015 mM). Complete-stack reproduction and independent external transfer remain open (validation §1.45).
- Breath H2/CH4 is useful as a broad fermentation/adherence proxy, not butyrate-specific quantification.
- One reviewed generic free-fatty-acid colorimetric protocol excluded acetic, propionic, and butyric acid. That representative mismatch does not exclude every future colorimetric chemistry.
Informs: quantification ladder · culture-supernatant transfer §1.31 · stool-stack transfer §1.45 · genotype-informed workflow
Detail: interpretive · experiment artifact · De Baere abstract scope; Gu full-text scope; matrix-specific qualification remains open.
comp-037 — C1-INH (SERPING1) EcN-Luminal Protease/Glycosylation Proxy — INVALIDATED MODEL (2026-05-17)¶
Question: Which sequence-filter, structure-confidence, glycosylation, and kinetic questions should be tested for a human C1-INH (UniProt P05155) payload produced by engineered E. coli Nissle 1917?
Verdict: Proxy only; empirical protease and glycosylation effects unresolved. The inherited LOW/RED labels used pLDDT as accessibility. Polypeptide-encoded serpin chemistry motivates testing an unglycosylated core, but exact-configuration folding, luminal stability, productive target engagement, and retained inhibition remain empirical gates.
Key findings: - Disulfide count grep-verified against UniProt FT DISULFID: exactly 2 disulfides (C123-C428, C130-C205) on the SV=2 entry. This is a sequence annotation, not folding evidence. - Candidate construct: serpin core aa 123–500. This exact boundary preserves the two annotated disulfides and RCL while making mucin-domain truncation directly testable; the proxy does not validate the construct. - UniProt features record N-glycans at 25, 69, 81, 238, 253, 272-variant, and 352 plus O-glycans at 47, 48, 64, 71, 83, 88, 92, and 96. The effect of removing the native N-terminal region remains empirical. - The retired five-enzyme filter panel did not model concentration, extended specificity, matrix, export topology, or kinetics and supplies no survival result. - Glycosylation precedents justify an unglycosylated test arm; they do not establish that an EcN-produced construct will fold or retain activity in the intended compartment.
Informs: complement-c5a-gout §9.8 · complestatin-bgc-lbp-feasibility-computational (comp-024 boundary) · engineered-lbp-chassis · hypotheses/H05 (sister-thread DAF SCR1-4 on koji) · current C1-INH evidence
Detail: interpretive · invalidated, non-runnable tombstone · Exact-configuration folding, stability, kinetics, and function remain open
comp-035 — Intra-articular Uricase H₂O₂ Reaction-Diffusion (3 Architectures) — NON-DECISION-GRADE PHASE-0 PRIOR (2026-05-16; downgraded 2026-07-14)¶
Question: What does a first-pass, well-mixed steady-state model predict for H₂O₂ handling by Pickering emulsion, uricase-catalase fusion, and free co-formulated catalase architectures?
Verdict: Non-decision-grade Phase-0 prior; no architecture is cleared or selected. The frozen v1 runs produced historical median/p95 values of 0.19/1.1 µM for Pickering, 0.034/0.20 µM for fusion, and 0.19/7.2 µM for free co-formulation. Those values describe the implemented assumptions only. They do not establish tissue safety because the steady-state threshold was unverified, loading and active-site accounting were unresolved, and local gradients and exposure time were not modeled.
Key findings: - Within the model, the low Pickering shell Damköhler result suggests that proximity alone did not drive the predicted bulk values; the alternative bulk-catalase explanation remains input-dependent and empirically unvalidated. - comp-035 does not establish a safe steady-state threshold, an architecture or chassis winner, or a basis for economics-driven selection. - Remaining gates are a matched reaction-site H₂O₂ time course; catalase activity, stoichiometry, retention, and diffusion; local exposure; and tissue safety.
Informs: chassis-pending-interventions §6 · gout-kill-chain-delivery-routes · delivery-route-matrix · engineered-koji-protocol
Detail: interpretive · frozen v1 artifact · Reviewed; non-decision-grade
comp-034 — Lactoferrin Inter-Lobe Linker Redesign — INVALIDATED MODEL¶
Question: If direct testing identifies a reproducible linker-associated failure, can the exact connector be redesigned while preserving lactoferrin fold and function?
Verdict: No candidate ranking survives. The model reused COMP-005's unverified protease-preference table as a biological cleavage axis. ProteinMPNN and Rosetta supplied additional model scores but did not validate that target. Cleavage values, GREEN/STRICT tiers, concordance claims, winners, and wet-lab priorities are invalid.
What survives: The exact connector remains a candidate engineering region only after WT fragment mapping or retained-function data identify a reproducible failure. A new design COMP must bind verified specificity and structural constraints before generating a matched diversity panel.
Informs: validation-experiments §1.10 · lactoferrin-protease-stability-computational · etc/bio-ai-tools · lactoferrin
Detail: interpretive · invalidated, non-runnable tombstone · Historical artifact retained in Git
comp-029 — Combined CP0 Scenario (RA + DAF SCR1-4) — INVALIDATED MODEL¶
Question: Could rosmarinic acid and active DAF SCR1-4 suppress MSU-associated complement activation more than either exact material alone?
Result boundary: Every numerical result, interval, category, co-localization claim, complementarity claim, and routing conclusion is invalid. The live corpus does not preserve those outputs; Git is the audit trail.
Next test: Once an active DAF preparation exists, compare vehicle, rosmarinic acid, DAF SCR1-4, and their combination in one matched MSU-associated complement assay. Measure C5a, C5b-9, DAF recovery, retained function, and surface association; do not route the experiment with an inferred accessibility coefficient.
Informs: complement-c5a-gout §9.5 · validation experiments · hypotheses/H05
Detail: current evidence · invalidated, non-runnable tombstone
comp-036 — Repeat-Dose Inhaled mRNA-IL-1Ra PK/PD (Receptor-Occupancy Framing) — YELLOW (2026-05-16)¶
Question: Does multi-administration inhaled mRNA-IL-1Ra dosing achieve clinically-meaningful sustained IL-1R1 receptor occupancy over the 72h acute gout flare window — how many doses, at what frequency, with what confidence bounds?
Verdict: YELLOW. Repeat dosing partially salvages comp-033's RED single-dose Cmax verdict, but the high-confidence GREEN bar (median 95% of the 0-72h flare window above 80% receptor occupancy AND p25 ≥ 50%) is NOT reached by any of three regimens tested (QD ×1–14, BID ×2–28, Loading 2× + QD-maintenance ×0–14). Modality viable but at the edge — wet-lab dose-finding needed.
Key findings: - Reframe from plasma Cmax-vs-anakinra (comp-033) to receptor-occupancy fraction over the 0–72h gout flare window — the clinically-relevant metric for a competitive antagonist. 80%-occupancy plasma threshold: median 73 ng/mL (p05–p95: 9–553). - Kd_nM is now the #1 sensitivity driver (Spearman ρ = −0.69), surfacing previously-implicit uncertainty. IL-1Ra-IL-1R1 Kd ~1 nM (Arend 1990 JCI, range 0.1–10 nM). - comp-033 RED single-dose verdict does not close the modality; repeat-dose receptor-occupancy is the right gate going forward.
Informs: chassis-pending-interventions §4 · inhaled-mrna-il1ra-pulse-computational · etc/open-enzyme-vision §10
Detail: interpretive · experiments/ · Complete v1
comp-033 — Inhaled mRNA-IL-1Ra Pulse Therapy Dose Modeling — RED on systemic anakinra-equivalent (2026-05-16)¶
Question: Does dose modeling show pulmonary IL-1Ra expression can plausibly reach anakinra-equivalent therapeutic exposure at currently-feasible inhaled-mRNA doses (4–24 mg per administration), and which inhaled-mRNA programs / CDMOs are forkable partners?
Verdict: RED on the systemic-anakinra-equivalent gate. Median predicted plasma Cmax 0.025 µg/mL = 1/60th anakinra (1.5 µg/mL); only p95 (0.28 µg/mL) approaches anakinra-trough (0.05 µg/mL). Reverse-dose calc: ~195 mg mRNA per administration to reach 0.5 µg/mL median; ~585 mg for full anakinra benchmark — both 8–25× the highest disclosed inhaled-mRNA clinical dose (24 mg, Translate Bio MRT5005). Verdict does NOT close the modality — comp-036 reframed to receptor-occupancy and pulls it back to YELLOW.
Key findings: - Dose-feasibility gap is the load-bearing finding; current inhaled-mRNA platforms are 1–2 orders below what plasma-Cmax-equivalence would require. - Three honest paths forward: (a) repeat dosing (→ comp-036), (b) reframe to local-pulmonary IL-1Ra exposure for inflammation-of-lung indications, © different target where lower Cmax suffices. - Partner-ID surface: Translate Bio (now Sanofi), Moderna, Arcturus, Ethris — none currently aimed at IL-1Ra; chassis-pending §4 stays active as a temporal-stack platform-positioning entry.
Informs: chassis-pending-interventions §4 · etc/open-enzyme-vision §10 · modality-chokepoint-matrix · delivery-route-matrix · repeat-dose-inhaled-mrna-il1ra-pkpd-computational (comp-036 reframe)
Detail: interpretive · experiments/ · Complete v1
comp-047 — ABCG2 Q141K Pharmacological-Chaperone Re-screen (real docking) — INCONCLUSIVE (2026-07-14)¶
Question: Does this static-receptor Vina configuration support a reproducible Q141K ABCG2 chaperone ranking after conservative known-ABCG2 exclusions?
Verdict: INCONCLUSIVE — no defensible docking-backed ranking. The corrected executable merge reports 0 yes and 1 uncertain row, vorinostat. Rosuvastatin is excluded because the FDA label identifies it as a BCRP substrate and the UniProt/DrugBank ABCG2 relationship set also flags it. Vorinostat's independent In Vitro Q141K-rescue precedent (Basseville 2012, PMID 22472121) does not validate the modeled pocket or make its docking row a wet-lab priority.
Key findings:
- The four CFTR correctors are cross-protein mechanism comparators, not validated ABCG2 chaperone positives. Their failure to earn a tier is a setup diagnostic, not evidence that ABCG2 cannot be rescued.
- The sensitivity run changed 2–7 of eight tracked positions across recorded perturbations. The base ordering is not robust for this setup; no absence-of-pocket conclusion follows.
- The exact receptor snapshot passes hash, residue-141, mutation-scope, and geometry checks with one declared symmetric PDBQT warning (SER655→UNK).
- The decisive next observation is validation §1.22, not another pass through the same static docking configuration.
Detail: interpretive · experiments/ · supersedes comp-032
comp-032 — ABCG2 Q141K Pharmacological-Chaperone Virtual Screen — ~~GREEN~~ SUPERSEDED (2026-05-16; superseded by comp-047 2026-07-14)¶
⚠️ Verdict RETRACTED. The GREEN below is a descriptor/class-prior heuristic whose separation of the CFTR cross-protein comparators was tautological. comp-047 replaced the class-prior ranking with a static docking test → INCONCLUSIVE. Retained as a frozen record; do not cite it as binding or rescue support.
Question: Is there an FDA-approved small molecule that binds ABCG2 Q141K's nucleotide-binding domain (NBD) and could rescue trafficking, or does the FDA-approved drug surface lack chaperone-active hits — requiring novel chemistry?
Verdict (retracted): ~~GREEN. Shortlist of 10 candidates passes the four-gate filter.~~ The shortlist is a prior-weighted descriptor ranking, not binding/rescue evidence; the "all four CFTR controls above decoy max" claim is a formula artifact of the 0.15 decoy prior (0.15^(⅕) = 0.684), not independent validation.
Key findings (historical, retracted as evidence): - Three chemistry classes scored ≥ 0.85 (CFTR correctors, tetramer stabilizers, bile-acid chaperones) — but the ranking is driven by the hand-assigned class prior, and ABCG2 inhibitors also scored high (the confound comp-047's Axis 2 targets). - Real-docking re-screen (comp-047) is now complete → INCONCLUSIVE, not the hypothesis-confirming step comp-032 anticipated. Do NOT treat the repurposing surface as empirically supported; do NOT queue a compounding-pharmacy conversation on this route.
Informs (now downgraded to hypothesis-only): chassis-pending-interventions §7 · abcg2-modulators §"Pharmacological-chaperone route" · compounding-pharmacy-track
Detail: interpretive (superseded) · experiments/ · superseded by comp-047
comp-031 — Dual-chassis EcN PDB + Uricase Additive SUA Prediction — INVALIDATED 2026-07-13¶
Original question: Would PDB-EcN plus luminal UOX produce additive ΔSUA and a PDB-derived butyrate→ABCG2/Q141K synergy?
Current verdict: INVALIDATED. The model inherited comp-019's unsupported saturation regime, transferred C. sporogenes butyrate yield into CBT2.0/EcN without product measurement, misattributed direct butyrate rescue to Basseville 2012, and added unmatched background butyrate to the combination arm. Its ΔSUA, competition, butyrate, Q141K-rescue, and two-strain engineering recommendations are retired.
Replacement work: comp-044 reopens the UOX regime; comp-046 conserves dietary fate, treats the endogenous side as a capture-fraction comparison, and supplies a conditional architecture boundary; validation §1.37 measures actual CBT2.0 products.
Detail: invalidated interpretive page · current invalidation record · obsolete model preserved only in Git history
comp-027 — Disulfiram Dose Modeling for GSDMD Blockade vs DER Ceiling — hypothesis-generator (2026-05-16; downgraded from YELLOW-leaning-GREEN by comp-review 2026-07-14)¶
Question: Is there a sub-AUD oral disulfiram dose window where plasma DSF engages GSDMD (CP6b pyroptotic-exit block) at a therapeutically meaningful level while plasma Me-DTC stays below the ALDH-inhibition threshold driving the disulfiram-ethanol reaction (DER)?
Verdict (downgraded to hypothesis-generator, comp-review 2026-07-14): a single strict-GREEN modeled point at 100 mg/day under current (unverified) Cmax + ALDH-calibration assumptions — NOT a validated "75–125 mg/day window" (those bounds were broadened from the single point, and 100 mg/day sits exactly on a hard-coded decision boundary). Hypothesis to test, not a dose recommendation. At 100 mg/d: ~57% GSDMD blockade (DSF Cmax ~0.4 µM) at ~40% ALDH inhibition (Me-DTC ~70 nM, right at Faiman DER hypotension threshold). Below 50 mg/d, GSDMD blockade drops <40%; above 125 mg/d, ALDH inhibition crosses DER threshold. Strict-GREEN at 100 mg/d under conservative cell-free EC50 anchor; cellular-preincub anchor extends GREEN down to 50 mg/d. Gates the 503A compounding-pharmacy disulfiram pathway.
Key findings: - Sub-AUD DSF is a selective GSDMD inhibitor, not a pan-NLRP3 inhibitor — the NLRP3-palmitoylation pathway (Xu 2024, 10 µM EC50) is NOT engaged at any sub-AUD dose. - DER threshold is the load-bearing ceiling; alcohol-abstention requirement is a compliance question for the 503A protocol. - Two EC50 anchors (cell-free vs cellular-preincub) bracket the GREEN window; cellular-preincub captures covalent-accumulation kinetics and is the more defensible anchor for chronic dosing.
Informs: compounding-pharmacy-track §6 · disulfiram · nlrp3-exploit-map CP6b
Detail: interpretive · experiments/ · Complete v1
comp-024 — Complestatin-Family BGC / LBP Proxy — INVALIDATED MODEL (2026-05-16)¶
Question: Can the historical hand-scored model establish a tractable LBP chassis or choose complestatin versus C1-INH as the next CP0 payload?
Verdict: Invalidated and retired. Hand-assigned factors, arbitrary color thresholds, and an incomparable nine-factor-versus-eleven-factor composite cannot establish feasibility, a chassis, comparator superiority, or portfolio priority.
Key findings: - Chiu 2001 reported a 48.7 kb cluster with 16 open reading frames and seven NRPS modules; Park 2016 reconstituted a 54.5 kb cluster in S. lividans and recovered monocyclic M55 and linear S56 from gene-deletion experiments. - The accessible Park primary record does not establish that M55 or S56 was inactive. - An exact host × oxygen-regime experiment must recover chemically identified, functionally active crosslinked product before the active-product-formation hypothesis advances. Delivery, access, safety, and priority require separate evidence. - C1-INH is a separate conjecture requiring exact-construct expression, folding, glycosylation-dependence, stability, function, and access measurements.
Informs: complement-c5a-gout §9.8 · engineered-lbp-chassis · modality-chokepoint-matrix
Detail: current evidence · invalidated, non-runnable tombstone
comp-030 — DAF SCR1-4 Cassette Ranking — INVALIDATED MODEL (2026-05-15)¶
Question: Which expression and processing configuration can produce intact, natively folded, functional DAF SCR1-4 in A. oryzae?
Verdict: Invalidated and retired. The candidate scores, promoted sets, codon preference, direct-secretion ranking, ESM2 pseudo-pLDDT inference, chaperone-load coefficients, and cross-target generalizations do not support candidate selection or experimental routing.
Key findings: - No numerical rank, count, preferred codon strategy, promoter, signal peptide, processing route, or fold/secretion conclusion survives. - Direct secretion and GlaA-KEX2 remain unranked candidate routes. Codon variants may be an independent factor, but the retired model supplies no preferred variant. - Compare exact constructs under matched conditions and measure expression, processing fidelity, native-fold attainment, intact secreted product, and retained complement-regulatory function.
Informs: validation-experiments §1.25 · hypotheses/H05
Detail: current evidence · invalidated, non-runnable tombstone
comp-022 — A. oryzae Uricase Cassette Ranking — INVALIDATED¶
Question: Could heterogeneous computational proxies identify an A. oryzae UOX cassette to advance?
Verdict: No; the ranking is invalidated and non-runnable. The CAI, RNA-structure, chaperone-load, promoter–signal-peptide, and ESM2 axes were not calibrated to one named biological outcome. No score, rank, tier, shortlist, winner, component preference, gene-synthesis recommendation, or expression/fold/secretion/function inference survives.
What survives: The declared 43,200-row enumeration is a historical inventory fact only. Direct secretion, GlaA-KEX2 processing, promoter, signal-peptide, codon, terminal, propeptide, and glycosylation choices remain unranked experimental factors. Exact constructs require matched expression, processing, localization, native-state, active-product, oxygen/peroxide, viability, and process-retention measurements.
Historical corrections: Four of four v1 cluster rows entered the retired v2 N-of-five ≥4 tier, but only one entered N-of-five =5; the strict tier included PTS1-blocked and unblocked routes. The file called esmfold_pLDDT.csv held rescaled single-pass ESM2 log probabilities, not ESMFold pLDDT. Q00511 residues 191–193 are NFS, not NSS.
Detail: current evidence boundary · invalidated, non-runnable tombstone · matched construct design · validation §1.5 and §1.33
comp-023 — cns1+cns2 Cordycepin-Burden FBA — INVALIDATED¶
Question: Could the encoded iWV1314 scenario establish the metabolic burden or multi-cassette compatibility of a cns1+cns2 cordycepin route?
Verdict: Invalidated, non-runnable artifact. The model converted source-reported batch-average productivity into a fixed continuous mmol/gDW/h demand using an assumed biomass density without time-resolved, condition-matched calibration. Together with unverified pathway assumptions, an artificial export bound, broken scenario boundaries, and separate capacity maxima mislabeled as yields, this prevents any burden, flux, breakpoint, product, feasibility, or compatibility conclusion.
What survives independently: Jeennor et al. directly demonstrated cns1+cns2-enabled cordycepin production in A. oryzae in their tested configuration (PMID 38071331). A cytosolic route avoiding direct ER-folding competition remains an unranked research conjecture; the exact four-arm isogenic product-and-cell-state experiment is specified on the evidence page.
Informs: cordycepin route and experimental gate · chaperone-orthogonal-stacking · validation-experiments
Detail: evidence page · non-runnable tombstone
comp-018 — Upstream Complement Modulator Sweep — INVALIDATED CATALOG (2026-05-17)¶
Question: Across all compound classes, which compounds have documented activity at upstream complement cascade nodes proximal to C5a generation, and which are gout-platform-relevant?
Verdict: Retired as a computational experiment. The script counted fields in a hand-curated catalog; it did not rerun searches, verify primary evidence, enforce translation review, or validate tiers. Cross-assay rankings, counts, dietary conclusions, chassis extrapolations, and engineering priorities are invalid.
Key findings: - Rosmarinic acid, luteolin, and Helicteres compounds remain assay-specific leads. The former COMP-020 is quarantined literature provenance and supplies no quantitative verdict. - Houttuynia cordata polysaccharides remain exact-material leads governed by the Houttuynia evidence page. - C1-INH remains a separate exact-configuration expression, folding, stability, function, and access question. - None of those leads inherits a rank, dietary inference, or chassis assignment from COMP-018.
Current evidence homes: legacy upstream-complement source inventory · Houttuynia · C1-INH
Detail: current routing page · invalidated, non-runnable tombstone · brief-contamination retrospective
comp-020 — QUARANTINED: legacy literature scan; no computational verdict (2026-05-08)¶
Historical question: Across upstream complement nodes, which compounds have documented direct modulator activity at a qualified ≤100 µM-equivalent threshold?
Disposition: Quantitative verdict invalid; artifact quarantined. This was a manual literature scan without an executable decision function or immutable source snapshot. It mixed heterogeneous assays, above-threshold observations, and mass-unit materials, so it did not establish a threshold-qualified hit set. Source-level observations may be reverified at their exact material, assay, unit, and provenance; no ranking, comparative potency, platform priority, dietary efficacy, gout-compartment potency, ChEMBL coverage rate, or exhaustive absence conclusion survives.
Key findings: - Rosmarinic acid, Helicteres lignans, luteolin/flavonoids, and exact polysaccharide materials remain assay-specific in-vitro leads, not a ranked class. - Search gaps are bounded records of what this scan did not find, not universal absence proofs. - The matched-material, matched-condition assay idea survives as a Research Conjecture; more literature arithmetic cannot replace that experiment.
Informs: complement-c5a-gout · hypotheses/H05 · tcm-gout-compound-triage-computational · medicinal-mushroom-compound-mapping-computational
Detail: current evidence boundary · quarantined legacy artifact
comp-001 — Uricase Shio-Koji Protease-Site Proxy (2026-05-05)¶
Question: Which Q00511 adjacent pairs match three fixed legacy preference filters, and what AlphaFold confidence surrounds each match?
Verdict: Proxy only; empirical risk unresolved. The analysis mapped adjacent-pair matches to unverified legacy filters and their pLDDT context. The filters are not established exhaustive protease-specificity rules, and pLDDT is model confidence rather than solvent accessibility or protease resistance.
Key findings: - The fixed filters returned 215 ALP, 97 NPr, and 44 acid-protease adjacent-pair matches. These counts describe the encoded filters, not demonstrated cleavage sites. - Q00511 has mean pLDDT 97.14 and minimum 80.50; no solvent-accessibility or SASA calculation was performed. - The analysis did not measure cleavage, retained activity, salt-conditioned protease behavior in the ferment, or any fermentation outcome.
Informs: validation-experiments §1.10 — supplies a fixed-filter and structural-confidence inventory while the empirical retained-activity assay remains the decision gate.
Detail: interpretive · experiments/ · Complete
comp-006 — DAF/CD55 Shio-Koji Protease Proxy (full ectodomain) — INVALIDATED (2026-05-05)¶
Question: What sequence-filter and AlphaFold-confidence prior does the inherited model provide for the DAF/CD55 soluble ectodomain (aa 35–353)?
Verdict: Proxy verdict invalid; empirical protease risk unresolved. The reported HIGH labels used pLDDT confidence as accessibility, and the maximum-risk construction does not discriminate among the retained sequence scopes. The Ser/Thr-rich stalk has lower AlphaFold confidence than SCR1–4 and remains a testable engineering liability, but the model does not establish solvent exposure, cleavage, degradation, or survival.
Key findings: - Removing the lower-confidence stalk is a falsifiable construct-design hypothesis, not a computationally validated stability improvement. - SCR1–4 has high AlphaFold confidence, but COMP-001 is not a validated protease-stability benchmark and cannot support a cross-payload survival comparison.
Informs: modality-chokepoint-matrix — Engineered soluble complement regulators row
Detail: interpretive · invalidated, non-runnable tombstone · Historical filter inventory only
comp-015 — T-axis Adjuvant Urate-Target Mapping — INVALIDATED¶
Question: Can mixed literature labels and estimated exposures identify the most gout-favorable androgen-active natural product?
Verdict: Invalidated for candidate comparison and decision use. The artifact mixed purified compounds, botanical extracts, related but non-identical quassinoids, animal and cell evidence, a null human safety-table outcome, and heuristic exposure calculations. No ranking, gout-direction verdict, evidence-cell comparison, or H-AN-02 adjudication survives. COMP-007 did not establish the achievable-concentration or IC50 thresholds that COMP-015 attributed to it; those inherited thresholds are withdrawn with the comparison method.
Source-specific leads retained: - Purified cordycepin lowered serum urate and renal URAT1 expression in hyperuricemic mice (Animal Model; PMID 29422889). - A 70% ethanol Eurycoma longifolia stem extract changed urate and transporters in hyperuricemic rodents, while eurycomanol-type compounds 4–7 inhibited hURAT1 uptake at 50 µM in cells; pure eurycomanone was comparatively low-activity in that assay (Animal Model + In Vitro; PMID 31920654). - Purified eurycomanol changed serum urate, clearance, hepatic PRPS expression, and transporters in hyperuricemic mice (Animal Model; PMID 34785103). Physta's human urate comparison was null.
Informs: exact-material evidence boundary · androgen-natural-modulation · wet-lab validation §2.8
Detail: invalidated, non-runnable tombstone · Git retains the retired artifact
comp-016 — T × Intestinal ABCG2 Suppression Evidence Mining — HISTORICAL BOUNDED SCAN (2026-05-07)¶
Question: Does primary literature support the load-bearing claim that androgens directly suppress intestinal ABCG2 expression at platform-relevant magnitudes?
Verdict: Direct androgen suppression was not demonstrated in the bounded scan. This narrow negative finding survives. COMP-016 did not resolve healthy-human intestinal sex stratification, and its ABCG2 attribution and magnitude claims are superseded by the primary-verified COMP-017 correction.
Key findings: - The scan did not identify a primary in-vivo demonstration of testosterone or androgen-receptor suppression of intestinal ABCG2. - The old Hoque 53%/88% intestinal values were search-summary tier and are not supported by the primary article or source-data workbook; use the 78% jejunal versus 44% renal Western comparison. - The 2021 paper is by Liu et al.; the 2023 Caco-2 paper is by Slepnev et al., not Klyushova et al. - Healthy-human baseline magnitude and functional urate-flux response remain open.
Informs: current ABCG2 evidence boundary · androgen-urate-axis · abcg2-modulators
Detail: historical bounded-scan interpretation · preserved COMP-016 artifact · current correction in COMP-017
comp-019 — Gut-Lumen Uricase × ABCG2 Genotype Stratification + Flux Model — INVALIDATED TOMBSTONE¶
Question: Can the gut-lumen uricase sink produce meaningful SUA reduction in non-Q141K males, or does it rely on Q141K-positive disease-state vulnerability?
Current verdict: INVALIDATED TOMBSTONE — non-runnable and not eligible for quantitative decision use. Phase A found no Q141K-stratified uricase clinical outcome in the sources searched for comp-019 as of 2026-05-08; this is not a universal absence claim. COMP-019's unconditional flat-dose classification is not robust to COMP-044's tested substrate-occupancy and finite-window diagnostics. COMP-044 supplies no replacement ΔSUA, dose, genotype order, physiological regime, efficacy model, topology/chassis selection, production-sufficiency target, or safety conclusion.
Key findings: - No Q141K-stratified uricase clinical outcome was identified in the comp-019 searched corpus as of 2026-05-08. - Physiological substrate, oxygen, access, survival, topology, and transit are now explicit gates. - Q141K remains a prospective stratification variable; adequate dose and responder ordering are open.
Replacement: comp-044 · comp-045 · validation §1.33
Detail: historical interpretation · hash-bound tombstone
comp-017 — Intestinal ABCG2 Sex-Difference Evidence Audit — DIRECT HUMAN BASELINE UNRESOLVED (corrected 2026-07-27)¶
Question: Do healthy-human intestinal ABCG2 distributions differ by at least 1.5-fold between sexes, and what source-bounded conclusions survive from the Hoque, Liu, Slepnev, and MacLean records?
Verdict: DIRECT HEALTHY-HUMAN BASELINE UNRESOLVED. No sex-stratified GTEx intestinal distribution or HPA intestinal protein values were extracted, so the prespecified 1.5-fold threshold was not tested. Healthy-rat, Q140K mouse disease-state, and nominal Caco-2 exposure results remain separate evidence contexts.
Key findings: - Hoque et al. 2020: 78% jejunal versus 44% renal Western-blot reduction in Q140K+/+ mice; the article separately reports reduced jejunal immunofluorescence signal. The historical 53%/88% sentence is not primary-source text. - Liu et al. 2021: nominal 100 µM estradiol benzoate increased Caco-2 ABCG2 mRNA at 48 hours without a dose-dependent response (In Vitro); physiological magnitude is not established. - Slepnev et al. 2023: nominal 1, 10, and 100 µM testosterone, estradiol, and progesterone increased Caco-2 ABCG2 (In Vitro, official English abstract tier). PXR/FXR inhibitor conditions do not directly exclude androgen-receptor involvement, and no serum multiplier is justified. - MacLean et al. 2008: no sex-specific difference reported in a healthy-rat intestinal transporter scan (Animal Model); this does not establish a human null.
Informs: current evidence boundary · historical COMP-016 scan · androgen-urate-axis · abcg2-modulators
Detail: evidence boundary · exact-snapshot artifact, outputs, and reviews
comp-014 — Medicinal Mushroom Compound × Chokepoint Mapping — INVALIDATED¶
Question: Could a bounded natural-product retrieval and target-mapping pipeline identify decision-relevant fungal leads?
Verdict: Invalidated and non-runnable. No compound row, count, target mapping, rank, safety flag, production claim, or advancement decision survives as current evidence.
What survives: Only design lessons and independently sourced research questions. Query labels are not provenance; retained-for-review is not safe; heterogeneous endpoints and polarities are not rankable; and bounded non-retrieval is not biological absence. Exact-material fungal mechanisms may be revisited through a new primary-source-anchored experiment.
Current research home: medicinal mushroom compound mapping
Detail: interpretive boundary · invalidated tombstone
comp-013 — TCM Gout Compound Triage — invalidated tombstone¶
Question: Which TCM compounds with documented gout indications have source-backed target evidence worth advancing to context-matched exposure and functional testing?
Verdict: No rank, viability label, exposure estimate, occupancy calculation, risk tier, dose conclusion, delivery route, or advancement decision survives. The retired model counted off-target evidence toward viability, failed to preserve target-effect polarity, and inherited invalid COMP-004 and COMP-007 shortcuts.
Key findings: - The nine names remain only an unranked historical search list; the names are not evidence. - Formula-level connections remain research leads but do not establish component, target, or synergy attribution. - COMP-049 is the separate Gate-1-only replacement design for a fixed mixed-source correction set. It has not run.
Informs: tcm-modern-rigor-intersection
Detail: interpretive · tombstone · Invalidated
comp-012 — DAF/CD55 SCR1-4 Truncated Shio-Koji Protease Proxy (2026-05-05)¶
Question: What sequence-filter and AlphaFold-confidence prior does the inherited model provide for stalk-truncated DAF SCR1-4 (aa 35–285)?
Verdict: Proxy verdict invalid; empirical risk unresolved. The reported LOW score used pLDDT confidence as an accessibility class. COMP-001 cannot validate that mapping or serve as a protease-stability benchmark. The stalk-truncation hypothesis remains separately reviewable under comp-012 and requires direct retained-activity testing.
Key findings: - Removing the lower-confidence stalk remains a testable construct-design hypothesis; the computation does not establish improved protease survival. - Expression, correct disulfide folding, retained complement function, processing stability, and mucosal delivery remain empirical gates.
Informs: complement-c5a-gout (CP0 evidence boundary) · hypotheses/H05 (stalk-truncation hypothesis) · modality-chokepoint-matrix (portfolio status)
Detail: interpretive · invalidated, non-runnable tombstone · Historical filter inventory only
comp-011 — C. utilis Uricase Cassette Compatibility — INVALIDATED MODEL (2026-05-05)¶
Question: What exact processing, folding, and secretion failures should be tested when wild-type P78609 or a patent-mutation proxy is placed in a Ward-style Aspergillus cassette?
Verdict: Invalidated and retired. The model lacked a planned CDS, mixed wild type with a synthetic mutation proxy, mis-mapped glycosylation coordinates, and converted unverified processing and cysteine proxies into categorical risk. No MODERATE verdict or exact-ALLN-346 interpretation survives.
Key findings: - UniProt P78609 identifies the wild-type sequence; the patent-disclosed mutation set does not establish the exact clinical ALLN-346 parent. - Retain wild type and a precisely defined mutation proxy only when a matched construct comparison is decision-relevant. - Produced termini, intact abundance, folding/assembly, compartment, and retained activity are the discriminating readouts.
Informs: uricase-variant-selection · validation-experiments §1.9
Detail: interpretive · invalidated, non-runnable tombstone
comp-010 — Cassette Compatibility for Dual-Cassette Koji Multi-Payload Configuration — INVALIDATED MODEL (2026-05-05)¶
Question: Does the uricase (Q00511) + lactoferrin (P02788) payload pair have cassette-design-specific issues (codon collisions, KEX2 geometry, secretion burden) that the Ward 1995 glucoamylase-KEX2 architecture won't handle?
Verdict: Invalidated and retired. The model inferred codon burden without an actual CDS, transferred KEX2-family specificity without an A. oryzae matrix, mis-mapped glycosylation coordinates, and converted bulk sequence proxies into a LOW secretion-burden verdict. No LOW, codon-collision, cleavage-certainty, host-capacity, or combined-cassette conclusion survives.
Key findings: - Ward 1995 remains a production precedent, not validation of the proposed sequence–junction–host configuration. - Internal sequence matches can nominate terminal mapping; they cannot predict cleavage or exclude a topology. - Build single-cassette controls first and measure transcript, produced termini, compartment, abundance, and retained function before a combined configuration.
Informs: validation-experiments §1.33 and §1.9 — exact-configuration measurement requirements; no topology selected
Detail: interpretive · invalidated, non-runnable tombstone
comp-007 — Food-Associated HDAC-Directed Candidate Screen — Invalidated¶
Question: Could heterogeneous HDAC and exposure evidence support a quantitative priority order for direct Q141K testing?
Verdict: Invalidated and retired. The model mixed assay types and analogical estimates, used arbitrary selectivity constants and 1 − oral bioavailability as an exposure surrogate, and did not use its stored concentration estimates. No rank, score, shortlist, causal HDAC-isoform assignment, HDAC6-centered safety inference, or advancement decision survives.
Key findings: - Seven compound names survive only as an unranked evidence inventory. - Selected pharmacological HDAC inhibitors provide an in-vitro Q141K-rescue precedent, but direct butyrate rescue, epithelial exposure, ABCG2-attributed urate flux, causal isoform mapping, and safety remain unestablished. - The possible combination of butyrate-associated endogenous ABCG2 induction and Q141K trafficking rescue is preserved as a Research Conjecture, not a COMP result.
Informs: validation-experiments §1.22 — direct, unranked candidate testing
Detail: current evidence · invalidated, non-runnable tombstone
comp-005 — Lactoferrin Shio-Koji Protease Proxy (2026-05-05)¶
Question: What sequence-filter and AlphaFold-confidence prior does the inherited protease proxy provide for human lactoferrin (P02788)?
Verdict: Proxy only; empirical protease risk unresolved. The reported HIGH/MODERATE labels used pLDDT confidence as an accessibility class and do not establish exposure, cleavage, degradation, or survival.
Key findings: - The exact inter-lobe connector is not a lower-confidence segment in the retired input. COMP-005 supplies no fragment-region priority. - Signal-peptide processing, glycosylation, context-specific protease activity, and retained function were not measured. - Inter-lobe-linker redesign remains a separate hypothesis that activates only if untargeted WT fragment mapping identifies a reproducible linker-associated failure.
Informs: validation-experiments §1.10 — both lactoferrin and UOX arms remain empirical feasibility gates
Detail: current evidence · invalidated, non-runnable tombstone
comp-004 — Supplement–ABCG2 assay-evidence audit — quantitative verdict invalid¶
Question: Do three cited ABCG2/BCRP interaction records—one each for quercetin, curcumin, and EGCG—support a quantitative prediction of intestinal urate-transport inhibition?
Verdict: No. The nominal bulk-concentration/drug-substrate-IC50 ratios, predicted inhibition percentages, and clinical-risk labels are invalid. The three records support ABCG2/BCRP interaction signals in different systems; they do not quantify intestinal urate transport.
Key findings: - Karibe supplies intestinal BCRP interaction evidence for curcumin with drug probes, not urate. - Cooray and Farabegoli supply non-intestinal drug-substrate signals for quercetin and EGCG. - The next gate is measured free parent/metabolites, ABCG2 protein and attribution, barrier integrity and viability, and basolateral-to-apical urate flux in an intestinal epithelial model.
Informs: validation-experiments §1.14 — direct context-matched urate-flux assay
Detail: interpretive · experiments/ · Complete
Planned Analyses¶
| ID | Scope | Primary informs | Priority |
|---|---|---|---|
| ~~comp-002~~ | Invalidated, non-runnable thermal/pH composite model; source-backed thermal sensitivity and the multi-day attrition conjecture now route directly to §1.10 measurement | §1.10 follow-up | Retired |
| ~~comp-003~~ | Reassigned 2026-05-05 → comp-005 (lactoferrin cleavage-site analysis) | — | ✓ Done as comp-005 |
| ~~comp-008~~ | Invalidated, non-runnable hand-scored payload rubric. Stable transformation + reporter expression is the common genetic-entry gate; native-pathway, uricase, lactoferrin, and CR1 tests remain separate unranked configuration questions. | engineered-lbp-chassis | Retired |
| ~~comp-009~~ | Invalidated, non-runnable target-site ranking. No guide, filter funnel, score, shortlist, GREEN verdict, accessibility, specificity, cross-species reuse, target-site-availability conclusion, H03 support, or P2-2 closure survives. | invalidated interpretation | Retired |
| comp-048 | Human proximal-tubule delivery-handle screen: identify surface-expression and topology candidates while keeping receptor identity, internalization, polarity, target coverage, and off-target expression as separate evidence gates | pre-run design | Gate 1 GO; not run |
| ~~Pre-numbering TCM triage attempt~~ | Later assigned to COMP-013; that attempt is now invalidated, and COMP-049 is a separate Gate-1-only replacement that has not run | — | Retired |
| ~~comp-021~~ | Invalidated, non-runnable mixed-tier assay-format model. Matched-format replication planning survives only as a conjecture; no quantitative range, candidate rank, or operative gut potency survives. | assay-format conjecture | Retired |
| ~~comp-022~~ | Invalidated, non-runnable cassette ranking. The 43,200-row enumeration survives only as historical inventory; every rank, tier, shortlist, winner, and component preference is retired. | evidence boundary | Retired |
| ~~comp-024~~ | Invalidated, non-runnable hand-scored model. Complestatin tailoring and C1-INH expression/folding/function require separate configuration-level tests. See Analyses above | — | Retired |
| ~~comp-023~~ | Invalidated, non-runnable FBA. Jeennor's direct A. oryzae production evidence and one ER-orthogonality conjecture survive independently; burden, flux, yield, breakpoint, and feasibility results do not. | cordycepin route | Retired |
| ~~comp-023 v2~~ | Deprioritized 2026-05-16 — koji-cordycepin removed from active stack (koji-endgame-strain §3.5) | — | Closed |
| ~~comp-025~~ | Closed after the independent portfolio decision to remove koji-cordycepin. Native cordycepin–pentostatin co-production survives only as an unranked ratio, exposure, and function measurement question. | — | Closed |
| ~~comp-026~~ | Deprioritized 2026-05-16 — multi-cassette induction interference moot for cordycepin; re-openable for future cytosolic third-cassette candidate | — | Closed |
| ~~comp-027~~ | Completed 2026-05-16; downgraded to hypothesis-generator 2026-07-14 — single strict-GREEN modeled point at 100 mg/d (not a validated 75–125 window). See Analyses above | — | ✓ Done |
| ~~comp-030~~ | Invalidated, non-runnable cassette-ranking model. Direct secretion and GlaA-KEX2 remain unranked configurations for a matched experiment. See Analyses above | — | Retired |
| ~~comp-029~~ | Invalidated, non-runnable toy scenario. The matched singleton/combination conjecture is grounded in independent single-arm evidence, not in the retired arithmetic. See Analyses above | — | Retired |
| ~~comp-031~~ | Not decision-usable — its flat-UOX, PDB-derived-butyrate, and compartment assumptions do not support the reported quantitative result or topology recommendation. See dual-chassis page, comp-044/046, and §1.34/§1.37. | — | ✓ Done |
| ~~comp-032~~ | Completed 2026-05-16 — ~~GREEN~~ SUPERSEDED by comp-047; verdict retracted because its CFTR-comparator separation was encoded by the class prior. See Analyses above | — | ✓ Done (superseded) |
| ~~comp-047~~ | INCONCLUSIVE. Corrected executable result: rosuvastatin excluded; vorinostat is one marginal row, not a docking-backed priority. The CFTR rows are cross-protein comparators, and recorded rank instability invalidates this static ordering without establishing that no rescue site exists. | Q141K trafficking + urate-flux assay §1.22 | ✓ Done |
| ~~comp-033~~ | Completed 2026-05-16 — RED single-dose Cmax-equivalent; reframed in comp-036 (YELLOW receptor-occupancy). See Analyses above | — | ✓ Done |
| ~~comp-036~~ | Completed 2026-05-16 — YELLOW repeat-dose receptor-occupancy framing; salvages comp-033 RED. See Analyses above | — | ✓ Done |
| ~~comp-037~~ | Invalidated, non-runnable proxy — exact-configuration folding, stability, kinetics, glycosylation effects, and function remain open. See Analyses above | — | Retired |
| ~~comp-038~~ | YELLOW — Tier 3 HPLC-UV culture-supernatant transfer candidate plus a separate Tier 2 electrochemical/ANN fecal candidate; no ready Tier ½ OE butyrate assay. See Analyses above | — | ✓ Done |
| ~~comp-039~~ | Reclassified as literature synthesis; no computational result. Four candidate-specific CFH-dependence conjectures survive, with no genotype-response direction. | CFH-dependence hypotheses | Reclassified |
| comp-040 | Wet-lab in-vitro CFH-replete/depleted/restored-serum MSU-crystal complement-activation assay — direct test of the candidate-specific CFH-dependence hypotheses for rosmarinic acid, luteolin, and qualified Houttuynia materials. The Houttuynia complement arm is independent of validation §1.30, which tests direct macrophage directionality; either may run when its exact material and assay capability are available, and failure in one does not adjudicate the other. | CFH-dependence hypotheses, Houttuynia, validation §1.30, gout-genetic-variants.md Category 5, complement-c5a-gout.md §6.3 | Blocked (OE wet-lab access and exact-material sourcing) |
| comp-041 | East Asian cohort feasibility scan for a Houttuynia exposure proxy × rs1061170 × incident gout cross-tab — assess whether a cohort records a usable exposure proxy and has adequate genotype/outcome support. Any resulting association would test that proxy in that population, not exact HCP biochemistry. | CFH-dependence hypotheses | Queueable (opens when an appropriate cohort and access path exist) |
| ~~comp-028~~ | Deprioritized 2026-05-16 — cordycepin-arm moot; general design-escape question non-load-bearing today; re-openable for future cytosolic third-cassette candidate | — | Closed |
Infrastructure proposals¶
comp-NNN verification agent (ClockBase hypothesis-then-verify pattern) — Planned (2026-05-08)¶
Every comp-NNN run produces output from a generation agent; add a second-pass verification agent (different vendor preferred per the multi-vendor heterogeneity discipline) that re-checks every load-bearing number (disulfide counts, residue indices, IC50/Ki, accession numbers, cohort sizes) against primary databases (UniProt, ChEMBL, PDB, PubMed) before commit. Sister discipline to the per-page Pre-commit verification gate (CLAUDE.md Rule 4) — same pattern at a different scope. Would have caught the 2026-05-06 DAF SCR1-4 disulfide hallucination at generation time. Cost ~$3–5 + 10–30 min per comp.
Detail: etc/autonomous-screening-methodology §"Hypothesis-then-verify pattern" · etc/manual-literature-mining §"Pre-commit verification gate" · operations/comp-018-vs-comp-020-retrospective
pcSec-class proteome-constrained A. oryzae GEM build — Planned (2026-05-14)¶
Layer secretion-pathway proteome-cost constraints on iWV1314 (Vongsangnak 2008): explicit PDI/calnexin/BiP saturation, signal-peptide processing capacity, KEX2 flux, and Sec61 throughput. This could evaluate a future secreted-protein cassette such as DAF SCR1-4 or an engineered C1-INH configuration. Complestatin NRPS biosynthesis is not a secreted-protein cassette and does not belong in this model. Any future cns1+cns2 analysis needs verified pathway boundaries and calibration against exact product and cell-state measurements; it must not reproduce COMP-023's retired scenario. Multi-week research project; not a single-subagent task.
Detail: chaperone-orthogonal-stacking · companion to verification-agent proposal (per-run vs per-strain infrastructure scopes)
How to add a new analysis¶
- Create
etc/experiments/comp-NNN-<slug>/withanalyze.py,inputs/,outputs/,README.md,inputs/provenance.md - Add an entry to the "Analyses" section above (compact format) or the "Planned Analyses" table
- Create
wiki/<slug>-computational.mdfor the interpretive page - Link from the relevant wet-lab experiment in
validation-experiments.md - Commit script + inputs + outputs together (outputs are version-controlled; they are the peer-reviewable artifact)