Skip to content

Medicinal Mushroom Compound × Chokepoint Mapping — Computational Analysis (comp-014)

Frozen analysis archived to ./etc/experiments/comp-014-medicinal-mushroom-compound-mapping/wiki-archive.md (248 lines). This wiki stub remains so cross-references resolve and the page stays discoverable. Computational analyses are write-once artifacts; the daemon does not need to re-read them on every sweep, so the long content lives next to the experiment that produced it at etc/experiments/comp-014-medicinal-mushroom-compound-mapping/.

Phase status

Phase What ran When Status
1 Scope: chokepoint targets, anchor species, toxicity filter, data-source inventory, multilingual literature plan 2026-05-06 Complete
2 Breadth aggregation across LOTUS + NPAtlas + KNApSAcK (NPASS / TCMSP / HIT unreachable from sandbox — documented gap with re-run plan) 2026-05-06 + 2026-05-17 Partial (3 of 6 planned DBs pulled)
3 Target mapping: ChEMBL activity.json per chokepoint UniProt → InChIKey intersection across 24 chokepoint targets, 26,830 unique ChEMBL molecules resolved, 323 (compound × chokepoint) empirical hits 2026-05-17 Complete for ChEMBL-empirical; SwissTargetPrediction prediction layer NOT RUN (sandbox-blocked, deferred)
4 (v2) Re-run chokepoint intersection on the unified compound table (9,778 compounds vs. the LOTUS-only 6,798 from Phase 2b) 2026-05-17 Complete
5 Multilingual literature deep-dive (CNKI / Wanfang / J-STAGE / KISS with two-model translation cross-check) per Phase 4 chokepoint-hit species — Queued
6 Per-compound triage (comp-013-style IC50 occupancy + composite scoring) on Phase 4 candidates — Queued

Phase 2 unified table — what's in / what's out

  • Total unique compounds (deduped by InChIKey): 9,778 (up from LOTUS-only 6,798)
  • By source: LOTUS 6,798 · NPAtlas 4,535 · KNApSAcK 20 (InChIKey-resolved subset of 398 raw records; PubChem InChIKey resolution wall-time-capped at 60 s per run)
  • Toxicity-pass: 9,747 compounds (after FDA GRAS / EFSA QPS / pharmacopoeia / clinical-trial inclusion + WHO 2022 / mycotoxin / Schedule I/II exclusion per inputs/toxicity-filter.json)
  • Hard-excluded: 31 compounds (mostly Aspergillus fumigatus + Stachybotrys chartarum + Penicillium expansum mycotoxin producers)
  • Anchor-species sanity check: all 18 anchor species from inputs/phase-5-anchor-species.json resolve in the unified table (passed)

Phase 2 database-coverage gap (NPASS / TCMSP / HIT / SwissTargetPrediction)

Three East-Asian-hosted compound DBs and the SwissTargetPrediction prediction service were not reachable from the comp-014 execution sandbox on 2026-05-17:

DB Endpoint probed Status Re-run plan
NPASS bidd.group/NPASS/ 200 once, then 000 (intermittent) Retry from non-sandboxed environment; prefer bulk-CSV download (NPASSv2.0_download_naturalProducts*.txt)
TCMSP old.tcmsp-e.com/tcmsp.php 502 Bad Gateway; alternate hosts unreachable Retry via BATMAN-TCM-2 endpoints (TCMSP folded into BATMAN-TCM-2) or Mendeley Data bulk archive
HIT hit2.badd-cao.net and badd-cao.net 000 timeout (both) Download HIT 2.0 Excel file per upstream paper PMID 35136829
SwissTargetPrediction swisstargetprediction.ch POST endpoint Not verified accessible from sandbox Batch the target-orphan compound subset (~98% of the table) through POST API in connected environment

The LOTUS-only Phase 2 outputs are not wrong, just incomplete. The unified table here is LOTUS + NPAtlas + KNApSAcK. The East-Asian-DB gap is documented and slated for re-run; do not over-interpret the absence of NPASS/TCMSP/HIT-only compounds as "no evidence."

Phase 3 — target mapping headline findings

323 empirical (compound × chokepoint) hits across 24 chokepoint targets, after toxicity filter. Coverage:

  • Compounds with ≥1 empirical chokepoint hit: 177 of 9,778 (1.81%)
  • Target-orphan compounds (no ChEMBL activity at any of the 24 chokepoints): 9,601 (98.19%) — Phase 3 SwissTargetPrediction layer is the load-bearing next step for closing this gap
  • Empty chokepoints (zero ChEMBL fungal-compound hits in this corpus): GLUT9, NLRP3, ASC, C5aR1, Lp-PLA2, KEAP1, OAT4, PINK1, PDI, PDIA3, TXN, TXNIP — twelve of 24 chokepoints have no fungal-source small molecule in ChEMBL with measurable activity at the target

Top empirical findings (potency-ranked, toxicity-pass only):

  1. Ganoderic acid H / Ganoderma lucidum — TNFα Kd = 2.45 nM (pChEMBL 8.61, ChEMBL1922178, 2014). The highest-potency single-compound × chokepoint hit in the entire breadth pass. Multiple stereoisomer / numbered-position variants of the same scaffold cluster at this Kd. Ganoderic acid D hits TNFα at Kd 8.39 nM (pChEMBL 8.08).
  2. Berkeleyamides A and D / Penicillium — CASP1 IC50 = 330 / 610 nM (pChEMBL 6.48 / 6.21, ChEMBL466565 / CHEMBL466747, 2008). The highest-potency direct caspase-1 fungal natural-product hits — distinct from the inflammasome-priming literature, this is direct effector-caspase inhibition.
  3. Berkeleyones A / B / C / Penicillium — IL-1β IC50 = 2.7 / 3.7 / 37.8 μM (pChEMBL 5.57-4.42, 2011). Modest potency but a direct IL-1β fungal-source hit.
  4. Quercetin / Agaricus — ABCG2 EC50 = 30 nM (pChEMBL 7.52, ChEMBL50, 2018). Plant-origin flavonoid that accumulates in mushroom substrates; the most-potent ABCG2 hit in the corpus.
  5. Ellagic acid / Penicillium / Phellinus — OAT1 IC50 = 270 nM (pChEMBL 6.57, 2005). Fungal-source organic anion transporter modulator.

Multi-chokepoint compounds (≥2 chokepoints hit, suggesting platform-level rather than single-target mechanism):

  • bois d'arc / morin (ChEMBL28626, Ganoderma): 4 chokepoints — ABCG2, CASP1, URAT1, XO
  • genistein (ChEMBL44, Cordyceps sinensis, Ophiocordyceps): 4 chokepoints — ABCG2, CASP1, PPARG, XO
  • Quercetin (ChEMBL50, Agaricus): 2 chokepoints — ABCG2, XO
  • Daidzein (Hericium, Cordyceps sinensis): 2 chokepoints — NRF2, XO
  • Disulfiram (Coprinopsis): 2 chokepoints — CASP1, NRF2

The "fungal-source" attribution for plant-origin flavonoids (quercetin, genistein, daidzein) reflects accumulation in mushroom substrates rather than fungal biosynthesis; see Phase 4 caveats.

Phase 3 implications for Phase 2 narrative

The Phase 2 LOTUS-only headline — Ganoderma applanatum 2,4-DAE as "strongest single-compound finding" — was correct as far as the LOTUS × PubMed pull went, but the unified ChEMBL intersection surfaces a substantially higher-potency direct hit: ganoderic acid H at TNFα Kd 2.45 nM. The 2,4-DAE finding is a (different, animal-model in vivo) Ganoderma applanatum hit at the urate axis (XO + URAT1); ganoderic acid H is a Ganoderma lucidum direct-binding finding at the cytokine axis (TNFα). Both belong in the candidate set; the breadth pass argues for taking Ganoderma triterpenoids more seriously across two distinct chokepoint axes, not just one.

The Phase 2 cordycepin → URAT1 finding (PMID 29422889, animal model) is preserved — ChEMBL has no direct cordycepin × URAT1 record. PubMed-derived in-vivo animal-model evidence and ChEMBL-derived in-vitro biochemical-assay evidence are complementary, not redundant.

Phase 3 NLRP3 "empty chokepoint" reversal — 2026-05-19 traditional-name re-scan

Material correction to the Phase 3 ChEMBL UniProt-join verdict. The Phase 3 intersection declared NLRP3 + ASC + Caspase-1 as "empty chokepoints" in fungi based on absent ChEMBL pure-compound activity entries. The 2026-05-19 mushroom traditional-name × NLRP3 re-scan revised this verdict:

  • ≥18 fungal sub-form × NLRP3-axis papers exist in PubMed under species-name + traditional-pathology framing
  • ≥5 at the gout indication itself (MSU + HUA rodent models)
  • Strongest single-species fit: Phellinus igniarius (桑黄) — covers XO + NLRP3 + URAT1 + bile-acid axes simultaneously across 4 independent papers
  • Cordyceps militaris (Wang 2023) puts NLRP3 — not URAT1 alone — as the primary anti-MSU mechanism head-to-head
  • Antrodia camphorata Antcin-H is an NLRP3-selective triterpenoid; Ganoderma lucidum S-GLSP (sporoderm-removed spore powder) and GLP4 (TBK1-binding pentapeptide) are distinct sub-fractions from the GLPP / triterpenoid bulk that hit the NLRP3 axis through different routes
  • Falsifying counter-finding: lentinan (L. edodes) was tested on MSU-arthritis and came back negative on NLRP3 — shiitake stays on AIM2 + eritadenine cardiovascular axes

The NLRP3 chokepoint is not fungal-empty; it was query-framing-empty. Phase 3's "empty" verdict reflected the limitation of seeding from ChEMBL pure-compound activity columns rather than from species-name + traditional-pathology PubMed framing — exactly the query-framing diagnosis the comp-018 Phase 2 upstream-complement-modulator sweep surfaced for the CP0 chokepoint. The pattern generalizes from complement (CP0) to NLRP3 (CP2-CP4), and is now expected to recur across other chokepoints unless seed-list construction is upgraded. See the 2026-05-19 query-framing retrospective audit for the broader audit and methodology fix.

Phase 3 ABCG2 retroactive addition — Poria cocos (2026-05-19)

Phase 4 v3 retroactive addition to the ABCG2 chokepoint hit list: Wolfiporia cocos (茯苓 Fu Ling) — Sun 2021 Front Pharmacol PMID 33651969. Hyperuricemic mouse model: both ethanol and water extracts significantly elevated intestinal ABCG2 mRNA + protein; water extract effect magnitude exceeded benzbromarone positive control (p < 0.01). Missed by Phase 3 ChEMBL UniProt-join because Poria triterpenes are outside ChEMBL's curated activity table. The mechanism is empirically uncharacterized (could be transcriptional / chaperone-class trafficking rescue / class I HDAC inhibition — none tested). Poria cocos is a canonical ingredient of Si Miao San (already in comp-013) but was implicitly included via the formula citation, not separately spawned as a compound source — the formula-completeness gap. See abcg2-modulators.md §"Tier 2 — Solid mechanism, modest evidence" for the full entry. (Animal Model; source: mushroom-hdac6-q141k-rescan-2026-05-19.md)

Phase 5a (partial) executed 2026-05-19 — NLRP3 axis only

The Phase 5 multilingual deep-dive was queued but never executed as a full chokepoint sweep. The 2026-05-19 traditional-name re-scan is a scope-limited partial Phase 5 — NLRP3 axis only:

  • Phase 5a (partial) executed 2026-05-19 — NLRP3 / IL-1β / Caspase-1 / ASC traditional-name deep-dive resolved per logs/mushroom-traditional-name-nlrp3-rescan-2026-05-19.md. The ChEMBL "empty chokepoint" verdict was query-framing-driven, not biology-driven. New medicinal-mushroom-complement-track species: Phellinus igniarius, Sanghuangporus vaninii, Inonotus hispidus, Antrodia camphorata. Falsifying finding: lentinan MSU-negative; shiitake stays on AIM2 / cardiovascular axes, not NLRP3.
  • Phase 5b (partial P0-1 remediation executed 2026-05-20) — CNKI route correction recovered Chinese-language urate leads for Ganoderma, Sanghuang/Phellinus, and Cordyceps militaris. First full publisher-page read: Xiong 2024 Biotechnology Bulletin whole C. militaris water extract in hyperuricemia rats (Animal Model; serum UA 281.62→93.27 µmol/L at 0.5 g/(kgd); URAT1/GLUT9 down, OAT1/ABCG2 up, hepatic XOD down; renal inflammatory-marker and microbiome-diversity effects). Sanghuang follow-up found open-access promotion-grade anchors outside the unresolved 2025 CNKI record: Hua 2023 Biomedicine & Pharmacotherapy S. vaninii ethanol extract in PO+adenine HUA renal-injury mice (125/250/500 mg/kg; URAT1/GLUT9 down, OAT1/OAT3/ABCG2 up; LD50 >5,000 mg/kg) and Sun 2022 Nutrients 1.5 g/kg S. vaninii / I. hispidus in YEP/OXO HUA mice plus MSU-induced acute gouty arthritis rats (serum UA/XOD and ankle-swelling/inflammatory-marker improvements). Phellinus follow-up found Chen 2023 Heliyon TFPI as an open-access Animal Model plus HK-2 renal-cell anchor (serum UA 105.0→82.3 micromol/L at 450 mg/kg; Cr 44.30→33.37 micromol/L; liver XOD 48.91→44.09 U/g prot; TLR4/NLRP3 down, ABCG2 protein up). The 2025 CNKI S. vaninii* UPLC-Q-TOF-MS record remains quarantined for full-text retrieval because the English/MT allopurinol-comparator phrase is ambiguous, and the 2025 CNKI Phellinus TFPI record remains quarantined for OAT1/ABCG2/URAT1 mRNA-versus-protein interpretation. Remaining Phase 5b scope still queued for URAT1, ABCG2, OAT1/OAT3, XO, GLUT9, C5aR1, DAF/CD55, Lp-PLA2, HDAC6, PPARG, Nrf2/KEAP1, PDI/PDIA3/TXN/TXNIP/GLRX across the recovered lead set. ABCG2 partial: Poria cocos added retroactively this scan. Source-read records: C. militaris, Sanghuang, Phellinus TFPI.

Where the analysis lives