2026-08-19 · cycle 0002 · UNREVIEWED
liver
Queued by a human. Must resolve against Open Targets before it is evidence.
Wipeout. Every candidate for liver died. That is a real result.
Check this yourself
- Disease id opens Open Targets, not our copy of it.
- Every trial is an NCT on ClinicalTrials.gov. Every paper is a PMID on PubMed.
- Challenger and critic saw the same evidence pack. Models are named below.
- Stamp is UNREVIEWED. No clinician signed this.
How this was run
- Disease id
- QUERY_liver on Open Targets
- Review
- UNREVIEWED. No oncologist signed this.
- Challenger
- gpt-5-6-luna
- Critic
- gpt-5-6-luna
- Evidence fetched
- 2026-08-19 16:48:54 UTC
Evidence pack both models saw
Same pack for challenger and critic. Links go to the public records, not to us.
Targets from Open Targets
- None returned
Known drugs
- None returned
Trials on ClinicalTrials.gov
- None returned
Papers on PubMed
- PMID 38447573 2024 · CellNeutrophil profiling illuminates anti-tumor antigen-presenting potency.
- PMID 39412509 2024 · Aliment Pharmacol TherSemaglutide 2.4 mg in Participants With Metabolic Dysfunction-Associated Steatohepatitis: Baseline Characteristics and Design of the Phase 3 ESSENCE Trial.
- PMID 36697706 2023 · Nat Rev Gastroenterol HepatolImmunology and immunotherapy of cholangiocarcinoma.
- PMID 37201670 2023 · J HepatolImmunobiology of cholangiocarcinoma.
- PMID 39658265 2025 · GutImmune checkpoint inhibitors and the liver: balancing therapeutic benefit and adverse events.
- PMID 40836099 2025 · NaturemAChR4 suppresses liver disease via GAP-induced antimicrobial immunity.
Entered01020304
- Atezolizumab plus bevacizumabkilled · mechanism
- Durvalumab plus tremelimumabkilled · mechanism
- Cabozantinibkilled · mechanism
- Gemcitabine plus cisplatin plus durvalumabkilled · mechanism
- Pemigatinibkilled · mechanism
- Ivosidenibkilled · mechanism
- Pembrolizumabkilled · mechanism
0 standing · 7 crossed out
Standing · 0
Wipeout. Nothing survived.
Killed · 7
Atezolizumab plus bevacizumab
killed · mechanismPD-L1 antibody plus VEGF antibody
KNOWNINFERRED
- Thesis
- For unresectable or advanced HCC, this is a conservative benchmark combining checkpoint blockade with angiogenesis inhibition. [KNOWN] Falsification: prospectively compare the combination with the active local standard in biomarker-defined HCC and stop if it fails to improve response durability or survival. [INFERRED]
- Mechanism
- Atezolizumab blocks PD-L1 signaling, while bevacizumab inhibits VEGF-mediated angiogenesis and may alter immune suppression in the tumor microenvironment. [KNOWN]
- Kill experiment
- In a randomized advanced-HCC comparison against the current local standard, require a prespecified improvement in overall survival or durable response; failure in the biomarker-defined population falsifies the proposed benefit.
- Why it died
- The pack defines only an unsolved broad liver query, not unresectable HCC. This is an established HCC regimen, not an unreviewed hypothesis. The causal chain stops at PD-L1/VEGF blockade: it does not identify a liver mutation or subtype, show that the subtype causes tumor formation, or connect treatment to the queried symptom. The pack itself reports no targets, drugs, or trials for QUERY_liver.
Durvalumab plus tremelimumab
killed · mechanismPD-L1 antibody plus CTLA-4 antibody
KNOWNINFERRED
- Thesis
- For advanced HCC, this is a conservative immune-combination comparator and tests whether dual checkpoint inhibition benefits patients who are unsuitable for or progress after other first-line approaches. [KNOWN] Falsification: evaluate the regimen in a randomized biomarker-stratified study and reject it if added CTLA-4 blockade increases toxicity without a durable clinical benefit. [INFERRED]
- Mechanism
- Durvalumab blocks PD-L1, while tremelimumab blocks CTLA-4-dependent T-cell inhibition. [KNOWN]
- Kill experiment
- Randomize biomarker-defined advanced-HCC patients to the combination versus an appropriate active comparator and kill the hypothesis if added CTLA-4 blockade produces no durable survival benefit while increasing clinically meaningful toxicity.
- Why it died
- This is an established advanced-HCC checkpoint regimen, while the queried disease is unspecified liver disease. No mutation, liver-cell lineage, tumor antigen, or causal path from liver disease to symptoms is supplied. The proposed mechanism is generic checkpoint pharmacology and cannot establish relevance outside HCC; the pack contains no supporting target or trial evidence for QUERY_liver.
Cabozantinib
killed · mechanismMultikinase inhibitor
KNOWNINFERRED
- Thesis
- Cabozantinib is a conservative later-line option for advanced HCC and a test of whether broader kinase suppression retains activity after immune-checkpoint exposure. [KNOWN] Falsification: measure benefit specifically after prior checkpoint therapy and discard the hypothesis if activity is no better than an appropriate approved comparator. [INFERRED]
- Mechanism
- Cabozantinib inhibits signaling through targets including VEGFR, MET, and AXL, pathways associated with angiogenesis, invasion, and treatment resistance. [KNOWN]
- Kill experiment
- In patients with advanced HCC previously exposed to checkpoint therapy, compare cabozantinib with an active approved later-line option and reject the hypothesis if prespecified survival or response endpoints are not superior.
- Why it died
- Cabozantinib is a known later-line HCC treatment, not an unreviewed liver hypothesis. VEGFR, MET, and AXL inhibition does not supply a causal chain from a defined liver mutation to a defined clinical symptom, and the candidate silently changes the query from liver disease to advanced HCC. No relevant target or trial is returned in the evidence pack.
Gemcitabine plus cisplatin plus durvalumab
killed · mechanismCytotoxic chemotherapy plus PD-L1 antibody
KNOWNINFERRED
- Thesis
- For advanced biliary tract cancer, including ICC, this is a conservative standard-of-care backbone that should be retained as the comparator for incremental immune or biomarker-directed strategies. [KNOWN] Falsification: test any proposed addition against this backbone and reject it if it does not improve clinically meaningful outcomes without unacceptable toxicity. [INFERRED]
- Mechanism
- Gemcitabine and cisplatin produce complementary cytotoxic stress, while durvalumab blocks PD-L1-mediated immune inhibition. [KNOWN]
- Kill experiment
- For advanced biliary tract cancer or ICC, test any proposed addition in a randomized comparison with this backbone and reject it if it fails to improve clinically meaningful survival or response without unacceptable added toxicity.
- Why it died
- The thesis switches from unspecified liver disease to advanced biliary tract cancer or ICC. This is a known treatment backbone, not an unreviewed hypothesis. Cytotoxic stress plus PD-L1 blockade gives no mutation-to-lesion-to-symptom chain for QUERY_liver, and the pack provides no target, trial, or disease-resolution evidence supporting the substitution.
Pemigatinib
killed · mechanismFGFR kinase inhibitor
KNOWNINFERRED
- Thesis
- For ICC with an activating FGFR2 fusion or rearrangement, pemigatinib is a biomarker-selected repurposing candidate whose activity should be confined to the molecularly defined subgroup. [KNOWN] Falsification: require an FGFR2 fusion or rearrangement before treatment and reject the hypothesis if matched tumors lack objective or durable responses. [INFERRED]
- Mechanism
- Pemigatinib inhibits FGFR1, FGFR2, and FGFR3 signaling; FGFR2 fusions or rearrangements can drive oncogenic signaling in a subset of ICC. [KNOWN]
- Kill experiment
- Require prospective confirmation of an FGFR2 fusion or rearrangement and test matched ICC for objective and durable responses; absence of activity in the molecularly matched group falsifies the proposed driver dependence.
- Why it died
- Pemigatinib is an established genotype-directed ICC therapy, not an unreviewed candidate for a broad liver query. The mechanism is only conditionally plausible: an FGFR2 fusion or rearrangement may drive a subset of ICC, but the pack neither establishes that alteration in the queried disease nor connects it through tumor biology to symptoms. Treating without the molecularly defined ICC diagnosis is a disease and delivery inference unsupported by the pack.
Ivosidenib
killed · mechanismMutant IDH1 inhibitor
KNOWNINFERRED
- Thesis
- For previously treated ICC carrying an IDH1 mutation, ivosidenib is a conservative genotype-directed candidate that tests whether suppressing the mutant metabolic driver delays progression. [KNOWN] Falsification: enroll only IDH1-mutant disease and reject the mechanism if target-matched tumors show no progression delay relative to control. [INFERRED]
- Mechanism
- Ivosidenib inhibits mutant IDH1 and is intended to reduce production of the oncometabolite 2-hydroxyglutarate. [KNOWN]
- Kill experiment
- Enroll only IDH1-mutant ICC and compare against an appropriate control; no prespecified progression-free or overall-survival benefit would falsify the proposed mutant-IDH1 dependence.
- Why it died
- Ivosidenib is a known treatment for previously treated IDH1-mutant cholangiocarcinoma, not an unreviewed hypothesis for liver disease generally. The chain stops at mutant-IDH1 inhibition and reduced 2-hydroxyglutarate; the pack does not establish an IDH1 mutation, ICC diagnosis, downstream differentiation effect, or symptom reversal in QUERY_liver. The absence of returned targets and trials further leaves the proposed transfer unsupported.
Pembrolizumab
killed · mechanismPD-1 antibody
KNOWNINFERRED
- Thesis
- Pembrolizumab is a biomarker-restricted candidate for liver cancers with MSI-H or dMMR biology, where enhanced mutation-associated antigenicity may support checkpoint sensitivity. [KNOWN] The long-shot extension is to identify immune-inflamed, antigen-presenting tumors without MSI-H or dMMR and test whether they show reproducible benefit. [INFERRED] Falsification: prospectively separate MSI-H/dMMR tumors from immune-inflamed biomarker-negative tumors and reject expansion if the latter lack durable responses. [INFERRED]
- Mechanism
- Pembrolizumab blocks PD-1 signaling and can restore antitumor T-cell activity when tumor immune evasion is checkpoint-dependent. [KNOWN]
- Kill experiment
- Prospectively separate MSI-H/dMMR-positive from biomarker-negative immune-inflamed liver tumors; retain the established biomarker-restricted hypothesis, but kill the extension if the biomarker-negative group lacks reproducible durable responses.
- Why it died
- Pembrolizumab is an established checkpoint drug, and the proposed MSI-H/dMMR indication is a tumor-agnostic biomarker hypothesis rather than evidence for the broad liver query. No MSI-H/dMMR status, liver tumor diagnosis, antigenic driver, immune infiltration, or causal connection to symptoms is present in the pack. The immune-inflamed extension is explicitly inferred and unsupported by the returned evidence.
