| EGFR — ALK | Competitive differentiation | B wins | gpt-5.6-luna | 94% | ALK offers stronger evidence of clinically meaningful differentiation over chemotherapy: in untreated advanced ALK-positive NSCLC, crizotinib improved progression-free survival and response rate versus platinum-pemetrexed chemotherapy (C2-EV_BIOMARKER_0002; C2-EV_CLINICAL_0001). Subsequent ALK inhibitors provide evidence of differentiated long-term and intracranial control (C2-EV_COMPETITIVE_0002) and broad activity against resistance mutations (C2-EV_COMPETITIVE_0004), while the documented fourth-generation pipeline shows an active differentiation landscape (C2-EV_COMPETITIVE_0005). The ALK case is not uniformly differentiated: there is substantial competition without direct head-to-head superiority among preferred agents (C2-EV_COMPETITIVE_0001), limited differentiation in one head-to-head comparison (C2-EV_COMPETITIVE_0006), and resistance or non-ALK-dependent progression can occur (C2-EV_BIOMARKER_0004). EGFR has strong mechanistic and tractability evidence, but the supplied dossier does not provide comparable direct chemotherapy-comparator or competitor-landscape evidence for the stated newly diagnosed setting. This verdict addresses scientific and clinical differentiation evidence, not market attractiveness. | - For EGFR: a randomized, newly diagnosed NSCLC comparison of an EGFR-directed small molecule against chemotherapy, with clinically meaningful endpoints such as progression-free survival, overall survival, or intracranial control.
- For EGFR: a documented current competitor pipeline and direct evidence of differentiation versus existing EGFR-directed treatments.
| |
| KRAS — EGFR | Patient stratification | A wins | gpt-5.6-luna | 78% | Both candidates support operational genomic stratification hypotheses: KRAS by mutation status and allele, particularly G12C/G12D, and EGFR by canonical activating mutations such as Exon-19 deletions and L858R. EGFR offers a highly specific molecular subtype and established biomarker-directed treatment logic, but the supplied high-quality evidence for KRAS more explicitly includes an independent prospective clinical sequencing cohort documenting distinct driver-defined subsets (C1-EV_GENETICS_0003), alongside prevalence and allelic-subtype data (C1-EV_GENETICS_0002; C1-EV_GENETICS_0004). This gives KRAS the clearer directly evidenced prospective feasibility signal, though only narrowly because the EGFR dossier also describes clinically actionable, testable subtype selection (C2-CAT-ONCO-BIO-036; C2-CAT-ONCO-BIO-046). | - Prospective, biomarker-stratified NSCLC treatment data directly comparing operational KRAS- versus EGFR-defined enrollment in the specified newly diagnosed, chemotherapy-backbone population, including assay thresholds and treatment outcomes.
| |
| ALK — KRAS | Modality–direction–exposure fit | A wins | gpt-5.6-luna | 88% | ALK has substantially stronger evidence for fit to a small-molecule modality: multiple approved oral ALK TKIs, regulatory validation, systemic and intracranial activity, and established biomarker-selection infrastructure. KRAS also has validated small-molecule tractability, but the supplied evidence is primarily mutation-specific, especially for G12C, and does not establish comparable fit for the broader KRAS G12C/G12D proposal in the specified tissue and cell context. Neither dossier supplies direct PK/PD or biodistribution evidence for the stated PDX setting, so the conclusion is driven mainly by modality precedent and demonstrated clinical target engagement rather than direct exposure data. | - Direct comparative PK/PD, target-engagement, exposure, and biodistribution data for each proposed small molecule in the specified NSCLC PDX model and mucus-producing glandular tumor cells, including duration of target suppression and adequacy of tissue penetration.
| |
| KRAS — ALK | Clinical validation | B wins | gpt-5.6-luna | 98% | ALK has substantially stronger indication-relevant clinical validation. The dossier contains multiple randomized phase 3 trials in biomarker-selected, previously untreated advanced ALK-positive NSCLC, including direct comparison with platinum-pemetrexed chemotherapy, superior progression-free survival and response outcomes, intracranial efficacy, durable follow-up, and regulatory approval extending into the adjuvant setting (C2-EV_CLINICAL_0001 through C2-EV_CLINICAL_0005; C2-EV_BIOMARKER_0002; C2-EV_TRACTABILITY_0005). These results use an established oral small-molecule modality and validated companion-diagnostic selection. KRAS has evidence of approved G12C inhibitors in NSCLC and clear small-molecule tractability (C1-CAT-ONCO-BIO-091b; C1-EV_TRACTABILITY_0001), but the dossier does not provide comparable randomized trial-level efficacy, dose/exposure, target-engagement, or endpoint detail for KRAS in the stated context. The ALK evidence is therefore clearly stronger, despite acknowledged limitations including lack of unadjusted overall-survival significance in PROFILE 1014 and resistance or heterogeneous dependence in some models. | - For KRAS, randomized NSCLC trial reports with molecule-specific dose, pharmacokinetic exposure, target-engagement/pharmacodynamic data, comparator outcomes, and prespecified endpoint results would most change the comparison.
- For KRAS, coded clinical trial-stoppage reasons and detailed evidence distinguishing target failure from inadequate exposure or molecule limitations are not supplied.
| |
| KRAS — MET | Human natural experiments | B wins | gpt-5.6-luna | 92% | MET has direct human pharmacological evidence matching the requested inhibitory direction: genotype-selected MET inhibition produced substantial and durable responses in MET exon 14-skipping NSCLC, including in treatment-naive patients, with a second inhibitor confirming activity (C2-EV_CLINICAL_0001, C2-EV_CLINICAL_0002, C2-EV_SAFETY_0002, C2-EV_SAFETY_0005). Toxicities are substantial but documented and monitorable, providing evidence relevant to tolerability (C2-EV_SAFETY_0004, C2-EV_SAFETY_0006, C2-EV_SAFETY_0007). KRAS has human somatic driver and germline RASopathy evidence, but no supplied human loss-of-function, protective-allele, or NSCLC pharmacological exposure evidence demonstrating both efficacy and tolerability of KRAS inhibition. The activating genetic direction for MET does not itself match inhibition, so the decision rests on the human pharmacological evidence rather than the genetic alterations. | - For KRAS: human pharmacological exposure data directly demonstrating efficacy and tolerability of KRAS inhibition in NSCLC, or a human loss-of-function/protective-allele phenotype whose direction matches inhibition.
- For MET: long-term exposure cohorts with mature tolerability data specifically separated from efficacy evidence.
| |
| KRAS — EGFR | Modality–direction–exposure fit | Abstained | gpt-5.6-luna | 96% | Both candidates have strong small-molecule modality precedent and in vivo tool-use evidence: mutant-selective KRAS inhibitors are clinically validated, and EGFR has multiple approved small-molecule TKI generations. However, the criterion specifically requires fit of modulation direction/depth, duration, tissue exposure, and delivery. The dossiers do not provide candidate-specific PK/PD, target-turnover, biodistribution, or tumor exposure data for the stated lung PDX and mucus-producing glandular-cell context. EGFR on-target toxicity is documented, but it does not establish superior modality–exposure fit. Ranking one candidate would therefore require supplying missing evidence or treating absence as evidence, so abstention is appropriate. | - Candidate-specific PK/PD modelling linking small-molecule concentration and target inhibition to tumor response in lung PDXs, including required depth and duration of KRAS or EGFR inhibition.
- Direct delivery, biodistribution, and tumor-versus-normal-lung exposure data in the specified mucus-producing glandular cell context.
- Target turnover and recovery-rate data establishing whether intermittent or sustained small-molecule exposure is required.
- Head-to-head evidence under the specified newly diagnosed, chemotherapy-backbone in vivo proof-of-concept setting.
| |
| MET — EGFR | Druggability and tractability | A wins | gpt-5.6-luna | 78% | Both targets have strong small-molecule precedent. EGFR has multiple approved TKI generations and available research tools, supporting broad chemical tractability (C2-EV_TRACTABILITY_0001; C2-CAT-ONCO-BIO-115). MET has more directly matched evidence for the specified NSCLC setting: genotype-selected oral MET inhibitors have regulatory approval and demonstrated clinical responses in NSCLC, including treatment-naive disease (C1-EV_TRACTABILITY_0002; C1-EV_TRACTABILITY_0003; C1-EV_TRACTABILITY_0004). MET also has in vivo PDX pharmacologic evidence (C1-EV_PERTURBATION_0002). The advantage is narrow because the dossiers do not provide comparative pocket-quality metrics, resolved structures, lung delivery data in the specified mucus-producing glandular cells, or protein-turnover evidence. MET tractability is also biomarker- and alteration-dependent, particularly for amplification (C1-EV_TRACTABILITY_0007), while EGFR inhibition has on-target toxicity that may narrow the window (C2-EV_SAFETY_0001). | - Head-to-head experimental structural and pocket-quality data for MET versus EGFR, including binding-site metrics, lung exposure or delivery in mucus-producing glandular cells, and comparative selectivity at clinically relevant concentrations.
| |
| KRAS — EGFR | Time to decisive experiment | Abstained | gpt-5.6-luna | 99% | Both candidates have evidence of available tool compounds and prior preclinical model use (C1-CAT-ONCO-BIO-010, C1-CAT-ONCO-BIO-115; C2-CAT-ONCO-BIO-010, C2-CAT-ONCO-BIO-115). However, the criterion specifically requires evidence about organisational readiness, capacity, cost, and historical cycle times. The dossiers do not provide comparable organisation-specific evidence, so neither target can be adjudicated as faster or lower cost without inference beyond the supplied evidence. | - Organisation-specific assay and reagent inventory, PDX model readiness, internal personnel and capacity, expected study timelines, study costs, and historical cycle times for comparable in vivo proof-of-concept experiments for each target.
| – |
| ALK — MET | Integrated therapeutic hypothesis | A wins | gpt-5.6-luna | 88% | ALK represents the stronger integrated hypothesis. It has a precisely defined oncogenic driver mechanism, selective perturbational dependency, a validated predictive biomarker, established oral small-molecule tractability, and multiple randomized studies showing superior systemic and intracranial outcomes over chemotherapy or earlier ALK inhibition in untreated ALK-positive NSCLC (C1-EV_CLINICAL_0001, C1-EV_CLINICAL_0002, C1-EV_CLINICAL_0004, C1-EV_BIOMARKER_0001, C1-EV_TRACTABILITY_0002). MET also has credible genotype-restricted clinical validation, especially for METex14, and direct PDX perturbation evidence (C2-EV_CLINICAL_0001, C2-EV_CLINICAL_0002, C2-EV_PERTURBATION_0002). However, its broader target hypothesis is weakened by strong assay and alteration dependence, limited activity in lower-level amplification, negative results for nonspecific MET selection, and a crowded METex14 competitive field (C2-EV_BIOMARKER_0004, C2-EV_CLINICAL_0006, C2-EV_COMPETITIVE_0005). ALK likewise has resistance, heterogeneity, and competitive limitations, but the depth and consistency of biomarker-linked clinical validation and CNS-active small-molecule efficacy provide the stronger overall target-disease-modality coherence. | - The Q1-Q19 criterion-weighted composite scores, required by the Q20 guardrail, are not supplied, so consistency between this integrated judgment and that composite cannot be reviewed.
- Direct comparative evidence of differentiation against currently preferred ALK or MET agents in newly diagnosed disease.
- Evidence specifically using mucus-producing glandular-cell models and the stated patient-derived xenograft context for both candidates.
| |
| MET — KRAS | Reversibility and therapeutic window | A wins | gpt-5.6-luna | 68% | MET has evidence for an approved oral small-molecule modality, clinically described as generally manageable, with established treatment discontinuation and monitoring practices. KRAS G12C intervention is explicitly covalent and irreversible, which increases persistence of target modulation and reduces reversibility if toxicity emerges, although mutant selectivity may spare wild-type KRAS. The MET window is not clearly broad—edema, ILD/pneumonitis, hepatotoxicity, and permanent discontinuation are documented—so the advantage is narrow. Direct comparative PK, exposure-response, and recovery data are absent. | - Comparative human exposure-response and dose-toxicity data, modality half-lives, duration of target engagement after dosing cessation, and documented recovery after interruption for both MET and KRAS interventions.
| - MET is clinically tractable in NSCLC through genotype-selected, oral small-molecule kinase inhibition of MET exon 14–skipping tumors.
- Capmatinib produced substantial responses in MET exon 14–skipping NSCLC, providing direct clinical proof of target tractability for this biomarker-defined subgroup.
- Manageable on-target profile
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Weist KRAS Mutationen auf, die neue medikamentös adressierbare Bindungsstellen schaffen?
- Mutant-selective inhibitors spare wild type
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Weist KRAS Mutationen auf, die neue medikamentös adressierbare Bindungsstellen schaffen?
|
| EGFR — MET | Reversibility and therapeutic window | Abstained | gpt-5.6-luna | 94% | Both candidates have small-molecule tractability evidence, but the dossiers do not provide the modality half-life, target-occupancy duration, pharmacodynamic recovery, or intervention-specific interruption/reversal data required to compare reversibility. Safety evidence is also not directly comparable: EGFR has low-confidence evidence of on-target rash and diarrhea (C1-EV_SAFETY_0001), while MET has evidence of a generally manageable profile but also substantial and potentially serious toxicities including edema, interstitial lung disease, hepatotoxicity, and treatment discontinuation (C2-EV_SAFETY_0001, C2-EV_SAFETY_0004, C2-EV_SAFETY_0006, C2-EV_SAFETY_0007). Under the guardrail, therapeutic window and reversibility must be judged together; the missing reversibility and exposure–response evidence prevents a defensible ranking. | - Candidate-specific exposure–response relationships and quantitative therapeutic-index data.
- Small-molecule pharmacokinetic half-life, target-occupancy duration, and pharmacodynamic recovery after interruption for each intervention.
- Direct evidence that adverse effects can be titrated, interrupted, or reversed, including outcomes after dose reduction or discontinuation.
- A comparable head-to-head or otherwise standardized assessment of reversibility and toxicity burden for EGFR versus MET inhibition.
| |
| MET — ALK | Clinical validation | B wins | gpt-5.6-luna | 94% | Both targets have indication-relevant clinical validation through biomarker-selected responses, approved small-molecule inhibitors, and evidence of on-target activity. MET has strong clinical evidence in MET exon 14-skipping NSCLC, including durable responses with capmatinib and traditional approval of capmatinib and tepotinib (C1-EV_CLINICAL_0001, C1-EV_CLINICAL_0002, C1-EV_CLINICAL_0003). However, the principal MET evidence is phase 2 and largely non-randomized, and validation is highly dependent on the specific alteration; lower-level MET amplification and nonspecific MET IHC were not reliably predictive (C1-EV_CLINICAL_0006, C1-EV_BIOMARKER_0004). ALK has stronger indication-relevant validation because it includes randomized phase 3 evidence against platinum-pemetrexed chemotherapy in untreated advanced ALK-positive NSCLC, with improved progression-free survival, response rate, symptoms, and quality of life (C2-EV_CLINICAL_0001, C2-EV_BIOMARKER_0002), followed by randomized phase 3 confirmation with alectinib, brigatinib, and lorlatinib showing systemic and intracranial benefit (C2-EV_CLINICAL_0002, C2-EV_CLINICAL_0003, C2-EV_CLINICAL_0004). Regulatory approval of ensartinib based on randomized phase 3 evidence further supports reproducibility across molecules (C2-EV_CLINICAL_0005). The lack of unadjusted overall-survival significance in PROFILE 1014 is a limitation, but substantial crossover makes this less informative about target validity (C2-EV_CLINICAL_0006). The ALK evidence therefore has the stronger trial design, comparator, endpoint breadth, treatment-naive population relevance, and cross-molecule reproducibility for the stated NSCLC inhibition context. | - Direct randomized comparative clinical evidence for MET inhibition versus chemotherapy or another active control in newly diagnosed, biomarker-selected NSCLC, with exposure-confirmed target engagement and outcomes specifically reported for mucus-producing glandular tumor cells.
- Direct comparative pharmacokinetic/exposure and dose-response data between the MET and ALK molecules in the stated indication and model context.
- Clinical validation specifically in the stated mucus-producing glandular-cell population and in a patient-derived xenograft-to-clinical translational bridge.
| |
| KRAS — MET | Reversibility and therapeutic window | B wins | gpt-5.6-luna | 68% | MET is supported as an oral, genotype-selected small-molecule intervention with a generally manageable on-target profile, although clinically important toxicities require monitoring and treatment modification. KRAS G12C inhibition is explicitly covalent and irreversible, which shifts risk toward reduced reversibility despite mutant selectivity and relative sparing of wild-type KRAS. The evidence therefore favors MET for the combined reversibility and therapeutic-window criterion, but only narrowly because direct MET reversibility, half-life, and comparative exposure-response data are not supplied. | - Direct comparative binding reversibility, pharmacokinetic half-life, exposure-response separation, washout kinetics, and clinical outcomes after interruption for MET and KRAS interventions.
| |
| ALK — KRAS | Human genetics and causal evidence | B wins | gpt-5.6-luna | 79% | KRAS has the stronger, though only narrow, human tumor-genomic evidence package: recurrent activating G12C/G12D mutations with an allelic series and concordant direction of effect, replicated in TCGA, an independent prospective MSK-IMPACT cohort, and a large advanced-NSCLC cohort (C2-EV_GENETICS_0001 through C2-EV_GENETICS_0004). ALK has strong direct causal evidence for EML4-ALK fusions and replication in tumor profiling (C1-EV_GENETICS_0001 through C1-EV_GENETICS_0003), but the supplied evidence is narrower and does not provide the requested population-genetic or QTL-based evidence. The comparison is limited because both dossiers rely primarily on somatic tumor-genomic observations rather than GWAS/fine-mapping or molecular-QTL colocalisation. | - Neither dossier supplies GWAS or fine-mapping, L2G-style variant-to-gene scores, molecular-QTL colocalisation, ancestry-stratified replication, or biobank phenotype analyses. Direct comparative evidence of these types would most change the answer.
| |
| MET — ALK | Safety and essentiality | Tie | gpt-5.6-luna | 84% | Both candidates have evidence supporting a therapeutic window: ALK has a viable adult knockout phenotype with specific reproductive and neuroendocrine effects, while MET has a reported manageable on-target profile. However, both small-molecule inhibitor classes also show clinically important toxicities, including hepatotoxicity, interstitial lung disease/pneumonitis, and embryo-fetal toxicity, with additional target-class-specific burdens for each. The dossiers lack direct human LoF, pan-essentiality, and systematic normal-tissue breadth data needed to establish a clear comparative advantage. The available safety evidence is therefore comparable rather than clearly favourable for either target. | - Human LoF data, pan-essentiality scores, and systematic normal-tissue expression breadth for both MET and ALK; target-specific separation of on-target toxicity from molecule- and modality-specific toxicity.
| - Manageable on-target profile
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- Favourable on-target safety
- Wild-type ALK appears to have limited essentiality for survival of adult normal tissues, although loss of ALK produces specific reproductive and neuroendocrine phenotypes in mice.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
|
| KRAS — MET | Modality–direction–exposure fit | B wins | gpt-5.6-luna | 78% | SECOND has the narrower but more directly documented modality–direction–exposure fit. MET exon 14 skipping removes the CBL-binding region, reduces receptor down-regulation, and produces sustained MET signaling that is directly inhibited by oral small-molecule MET TKIs; MET-dependent NSCLC PDX responses and a validated pMET:total-MET pharmacodynamic assay support target engagement and duration assessment (C2-EV_MECHANISM_0002, C2-EV_PERTURBATION_0002, C2-EV_BIOMARKER_0005). Clinical activity in METex14-selected NSCLC further supports that inhibition depth is achievable with this modality (C2-EV_CLINICAL_0001). FIRST is also highly tractable for small-molecule inhibition, including mutant-selective covalent inhibitors and PDX use (C1-EV_TRACTABILITY_0001, C1-CAT-ONCO-BIO-010), but the dossier provides less direct evidence for PK/PD, biodistribution, and duration in the specified tissue and cell context. The margin is narrow because neither dossier supplies the requested direct exposure or delivery data, and MET expression can be heterogeneous or absent in some METex14 tumors (C2-EV_TISSUE_0006). | - Direct comparative PK/PD, lung-tumor biodistribution, and exposure–response data for the proposed small molecules in the specified mucus-producing glandular-cell PDX context.
- Candidate-specific target-engagement and duration-of-inhibition measurements in the stated in vivo proof-of-concept model.
- Evidence that either target is present and functionally relevant specifically in the specified mucus-producing glandular cell population rather than predominantly in tumor cells.
| |
| MET — KRAS | Time to decisive experiment | Abstained | gpt-5.6-luna | 97% | The dossiers provide partial evidence of tractability and prior in vivo model use for MET (C1-EV_PERTURBATION_0002) and available KRAS tool compounds and PDX precedent (C2-EV_TRACTABILITY_0001; C2-CAT-ONCO-BIO-010), but they do not report comparative assay/reagent readiness, internal capacity, historical cycle times, or costs. Therefore, the evidence does not support a reliable faster/lower-cost determination for either candidate. | - Comparative assay and reagent availability, including actual procurement or deployment readiness
- Comparative access to suitable NSCLC patient-derived xenograft models and their operational readiness
- Internal capability, staffing, throughput, and capacity for the in vivo proof-of-concept studies
- Historical cycle times from study initiation to interpretable in vivo results
- Per-candidate study cost and resource requirements
| – |
| KRAS — EGFR | Biomarker and target engagement | B wins | gpt-5.6-luna | 84% | Both candidates have credible genomic patient-selection strategies and regulatory precedent. KRAS has recurrent NSCLC mutations, an established G12C-selected population, approved mutation-directed inhibitors, and a proximal KRAS-GTP pharmacodynamic readout (C1-EV_GENETICS_0001; C1-CAT-ONCO-BIO-091b; C1-CAT-ONCO-BIO-099). EGFR provides a somewhat broader and more established clinical biomarker framework in NSCLC: replicated causal Exon-19 deletion and L858R biomarkers, routine biomarker-guided indication-specific treatment precedent, approved mutation-directed TKIs, and a direct pEGFR target-engagement readout with downstream efficacy-linked use (C2-EV_GENETICS_0001; C2-CAT-ONCO-BIO-046; C2-CAT-ONCO-BIO-091b; C2-CAT-ONCO-BIO-099; C2-CAT-ONCO-BIO-036). The advantage is narrow because the dossiers do not provide direct clinical serial-biomarker data or explicit assay validation details for either candidate, and KRAS-G12C itself also has strong regulatory validation. | - Direct head-to-head clinical evidence linking serial tissue or circulating biomarker measurements, target-engagement/pharmacodynamic changes, and early efficacy outcomes in newly diagnosed NSCLC for each target.
- For both candidates, explicit documentation of the clinical assay platform, analytical validation, specimen requirements, turnaround time, and regulatory companion-diagnostic status.
| |
| MET — EGFR | Human genetics and causal evidence | A wins | gpt-5.6-luna | 82% | FIRST has the stronger documented causal human-genetic evidence. MET alterations are directly assigned to MET through recurrent somatic exon-14 splice-site variants, a large allelic series with 126 distinct variants, cis-acting splice effects, functional activation, and replication across large lung-cancer cohorts (C1-EV_GENETICS_0002, C1-EV_GENETICS_0003, C1-EV_GENETICS_0004). SECOND has a concise claim that EGFR exon-19 deletions and L858R are causal with strong replication and directional effects (C2-EV_GENETICS_0001), but it supplies substantially less primary genetic detail. Neither dossier provides the requested GWAS/fine-mapping, L2G, molecular-QTL colocalisation, rare-variant burden, or ancestry-diverse replication evidence, so the margin is narrow. C1-EV_GENETICS_0007 limits generalization of MET as a broad mechanism but does not overturn the direct causal evidence for the defined MET-altered subset. | - For both candidates: ancestry-diverse GWAS or biobank replication with fine-mapped variant-to-gene assignment, L2G-style evidence, disease-relevant molecular-QTL colocalisation, and rare-variant burden analyses.
- For EGFR: detailed primary-cohort evidence quantifying the replication, allelic series, and direction of effect underlying the claim in C2-EV_GENETICS_0001.
- For MET: independent ancestry-diverse genetic replication beyond recurrent somatic tumor-sequencing cohorts.
| |
| MET — ALK | Human natural experiments | B wins | gpt-5.6-luna | 74% | Both candidates have human pharmacological exposure evidence consistent with efficacy under inhibition and clinically characterized tolerability. MET is supported by responses to selective inhibitors in MET exon 14-altered NSCLC, including treatment-naive patients, but the evidence is primarily phase 2 and accompanied by substantial edema, hepatic, pulmonary, and other toxicities (C1-EV_SAFETY_0002, C1-EV_SAFETY_0005, C1-EV_SAFETY_0004). ALK has broader and higher-level human exposure evidence: randomized trials show efficacy over chemotherapy and durable long-term benefit, including the 7-year CROWN follow-up, while safety is clinically characterized though not benign (C2-EV_CLINICAL_0001, C2-EV_CLINICAL_0004, C2-EV_SAFETY_0003, C2-EV_SAFETY_0005). The genetic evidence for both is somatic gain-of-function tumor biology rather than lifetime human loss-of-function, so it does not establish a true protective natural experiment and cannot be treated as equivalent to time-limited pharmacological inhibition. On the supplied human pharmacological evidence, ALK is better supported, but only narrowly because neither dossier supplies the strongest specified natural-experiment evidence. | - Neither dossier provides a human germline loss-of-function, protective-allele, or Mendelian-disease natural experiment directly linking target inhibition to reduced NSCLC risk with long-term tolerability.
- A directly comparable long-term drug-exposure cohort with standardized tolerability outcomes for both targets would most affect the narrow ranking.
| - Lorlatinib provides durable long-term clinical benefit in treatment-naive advanced ALK-positive NSCLC, with sustained systemic and intracranial disease control.
- ALK is clinically validated as a predictive therapeutic biomarker in advanced NSCLC by randomized phase 3 evidence showing superior efficacy of ALK inhibition over platinum-pemetrexed chemotherapy.
- Clinical responses to ALK inhibition provide strong evidence that ALK rearrangements are tumor-essential drivers in ALK-positive NSCLC.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- MET exon 14–altered NSCLC shows clinically meaningful dependence on MET signaling, supporting MET as an essential therapeutic driver in this molecular subset.
- A second selective MET inhibitor confirms both target dependence and a clinically relevant toxicity burden in MET exon 14–skipping NSCLC.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
|
| ALK — EGFR | Disease biology and mechanism coherence | A wins | gpt-5.6-luna | 90% | FIRST has the more coherent and experimentally supported inhibition mechanism. ALK rearrangements are supported as causal human NSCLC alterations, with transforming activity, defined downstream MAPK signaling, tumor-cell localization, and selective ALK knockdown effects in ALK-positive but not ALK-negative NSCLC models (C1-EV_GENETICS_0001, C1-EV_GENETICS_0002, C1-EV_MECHANISM_0003, C1-EV_MECHANISM_0004, C1-EV_PERTURBATION_0001, C1-EV_TISSUE_0001, C1-EV_TISSUE_0002). The mechanism is not uniform: resistance and variant-specific dependence weaken universal claims, but they do not negate the defined causal model (C1-EV_MECHANISM_0006, C1-EV_MECHANISM_0007). SECOND also has evidence that activating EGFR mutations are causal and produce ligand-independent kinase signaling into MAPK and PI3K pathways (C2-EV_GENETICS_0001, C2-EV_MECHANISM_0001), but much of the supporting material is catalogue-level synthesis rather than independent direct perturbation evidence. EGFR bypass and compensatory signaling further qualify the inhibition mechanism (C2-CAT-ONCO-BIO-004, C2-CAT-ONCO-BIO-013). The judgment concerns mechanistic coherence, not druggability; clinical tractability evidence was not used as a basis for the verdict. | - Independent, direct EGFR loss-of-function or pharmacologic inhibition studies in NSCLC models specifically representing the stated mucus-producing glandular-cell context, with matched human pathology or PDX evidence.
| |
| EGFR — KRAS | Clinical validation | A wins | gpt-5.6-luna | 78% | FIRST has stronger indication-relevant clinical validation because the dossier describes EGFR as a clinically validated NSCLC target with multiple approved small-molecule TKI generations and established targeting of common activating EGFR mutations, including exon-19 deletions and L858R (C1-CAT-ONCO-BIO-091b; C1-CAT-ONCO-BIO-088). SECOND also has meaningful clinical validation, but it is narrower: the supplied evidence specifically supports approval of mutant-selective KRAS-G12C inhibitors in previously treated KRAS-G12C NSCLC, rather than the broader KRAS-G12C/G12D context (C2-CAT-ONCO-BIO-091b; C2-CAT-ONCO-BIO-088). The margin is narrow because neither dossier supplies the trial-level dose/exposure, design, endpoint, population-selection, target-engagement, or stoppage data needed for a robust molecule-quality comparison, and neither directly matches the newly diagnosed, chemotherapy-backbone context. | - Indication-matched trial-level data for newly diagnosed EGFR-mutant versus KRAS-mutant NSCLC, including molecule identity, dose, exposure, confirmed target engagement, trial design, comparator, biomarker-selected population, and clinically relevant endpoints.
- Coded trial-stoppage reasons and completed-trial results for both targets in the specified lung NSCLC context.
- Direct clinical validation in the specified mucus-producing glandular-cell and chemotherapy-backbone context.
| |
| EGFR — MET | Safety and essentiality | B wins | gpt-5.6-luna | 68% | MET has the more favourable anticipated on-target safety profile, but only narrowly. Its dossier provides direct clinical evidence of a generally manageable on-target profile and biomarker-restricted dependence, although substantial class toxicities—including edema, interstitial lung disease/pneumonitis, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity—remain important liabilities (C2-EV_SAFETY_0001, C2-EV_SAFETY_0002, C2-EV_SAFETY_0004, C2-EV_SAFETY_0006, C2-EV_SAFETY_0007). EGFR has stronger evidence of broad physiological epithelial importance and severe knockout phenotypes affecting skin, lung, and gastrointestinal tissues, together with reported on-target dermatologic and gastrointestinal toxicity (C1-CAT-ONCO-BIO-098, C1-CAT-ONCO-BIO-102, C1-CAT-ONCO-BIO-114, C1-EV_SAFETY_0001). MET is incompletely characterized by human LoF, pan-essentiality, and normal-tissue expression evidence, so this is not a clear or broad safety advantage; the conclusion is driven by the available clinical safety and biomarker-restricted essentiality evidence rather than by absence of MET toxicity reports. | - For MET: human loss-of-function data, pan-essentiality scores, genome-wide normal-tissue expression breadth, and systematic normal-tissue knockout phenotypes
- For both targets: a directly comparable adult human genetic or tissue-specific therapeutic-window assessment in lung mucus-producing glandular cells
| - Manageable on-target profile
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- On-target dermatologic and GI toxicity
- Welcher Phänotyp tritt auf, wenn EGFR deletiert wird?
- Welche Rolle spielt EGFR in gesunden Zellen/Geweben?
- Welche Gewebe oder Organsysteme sind bei der Maus nach dem Knockout des EGFR-Gens betroffen?
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- On-target dermatologic and GI toxicity
|
| KRAS — MET | Perturbational evidence | B wins | gpt-5.6-luna | 92% | MET has the stronger reproducible PDX perturbational evidence. Pharmacologic MET inhibition produced antitumor responses across six MET-amplified lung-cancer PDX models, and a separate set of six patient-derived EGFR-mutant/MET-amplified models showed MET-governed signaling and sensitivity to MET inhibition in vivo (C2-EV_PERTURBATION_0002, C2-EV_PERTURBATION_0003). Genetic MET knockdown independently showed selective growth inhibition, arrest, and apoptosis in MET-amplified NSCLC cell lines (C2-EV_PERTURBATION_0001), supporting biomarker-restricted on-target dependence. The evidence is not universal: effects were largely confined to molecularly defined subsets (C2-EV_PERTURBATION_0004), and CRISPR knockout in A549 cells did not impair intrinsic proliferation although it affected dissemination-related phenotypes (C2-EV_PERTURBATION_0005). KRAS has supplied statements that tool compounds were used in PDX models, but the PDX studies and results are not comparably explicit or reproducibly characterized in the dossier (C1-CAT-ONCO-BIO-010, C1-CAT-ONCO-BIO-117b). | - For KRAS, explicit, independently replicated PDX perturbation studies with defined KRAS-altered NSCLC models, intervention details, and disease-relevant response measurements; the supplied KRAS PDX evidence is generic or insufficiently specified.
| |
| ALK — KRAS | Disease biology and mechanism coherence | A wins | gpt-5.6-luna | 95% | ALK has the more coherent inhibition mechanism for NSCLC: a recurrent human EML4-ALK fusion is shown to transform cells and is confirmed in human lung tumors, with defined constitutive kinase signaling and downstream MAPK dependence (C1-EV_GENETICS_0002, C1-EV_GENETICS_0003, C1-EV_MECHANISM_0003, C1-EV_MECHANISM_0004). Direct ALK knockdown selectively impaired viability and clonogenic growth in ALK-rearranged NSCLC models versus ALK-negative controls, providing experimentally causal support for the direction of modulation (C1-EV_PERTURBATION_0001). The ALK mechanism is not uniform across all tumors or fusion variants, as shown by resistance and heterogeneity evidence (C1-EV_MECHANISM_0006, C1-EV_MECHANISM_0007, C1-EV_PERTURBATION_0004, C1-EV_PERTURBATION_0005, C1-EV_PERTURBATION_0006), but these qualify the scope rather than eliminate the causal model. KRAS has strong human-genetic and pathway plausibility in NSCLC (C2-EV_GENETICS_0002, C2-EV_GENETICS_0003, C2-EV_MECHANISM_0001), but the dossier supplies substantially less direct, high-quality NSCLC-specific perturbational evidence establishing that inhibiting KRAS itself causally alters tumor pathophysiology. The decision therefore rests on mechanistic and perturbational coherence, not on druggability or clinical tractability. | - A direct, well-controlled KRAS inhibition or allele-specific genetic perturbation study in KRAS-mutant NSCLC models, ideally including rescue or pathway-level causal validation and comparison with KRAS-wild-type controls.
- Human NSCLC pathology or spatial evidence specifically localizing the relevant KRAS-mutant causal state to the disease-relevant tumor cell population.
| |
| MET — ALK | Patient stratification | B wins | gpt-5.6-luna | 84% | Both targets support prospectively testable genomic selection. METex14 skipping has regulatory approval, FDA-approved testing, and prospective clinical response evidence, but MET stratification is complicated by alteration-specific performance, amplification thresholds, assay discordance, and intratumoral heterogeneity. ALK provides the clearer operational hypothesis: ALK rearrangement is a defined mechanistic subtype, is selected with authorized companion diagnostics, and has been prospectively tested in randomized trials against chemotherapy and in subsequent-line biomarker analyses. The ALK hypothesis is therefore clearer, although the margin is narrow because METex14 is also clinically validated and prospectively actionable. | - A direct prospective comparison of ALK-rearrangement and METex14 selection using equivalent trial designs and diagnostic workflows would most change the answer.
| |
| KRAS — EGFR | Integrated therapeutic hypothesis | A wins | gpt-5.6-luna | 78% | FIRST is stronger overall for a differentiated small-molecule hypothesis because the dossier combines recurrent human NSCLC driver evidence with clinically validated mutant-selective covalent KRAS inhibition and evidence that wild-type KRAS can be spared, supporting a potentially differentiated therapeutic window (C1-EV_GENETICS_0002, C1-EV_GENETICS_0003, C1-EV_TRACTABILITY_0001, C1-EV_SAFETY_0001). EGFR has at least comparable causal driver evidence and greater clinical maturity, including multiple approved TKI generations and a specific NSCLC indication (C2-EV_GENETICS_0001, C2-EV_TRACTABILITY_0001, C2-CAT-ONCO-BIO-046), but the supplied safety evidence identifies on-target dermatologic and gastrointestinal toxicity from wild-type EGFR inhibition (C2-EV_SAFETY_0001), which weakens differentiation. The margin is narrow because neither dossier directly validates the specified cell type, newly diagnosed setting, chemotherapy combination, or exact PDX model; the KRAS advantage is therefore an integrated modality/differentiation judgment rather than direct evidence in the full stated context. | - Direct head-to-head evidence in the specified mucus-producing glandular-cell NSCLC context, including matched genomic biomarker-defined models.
- In vivo PDX proof-of-concept data specifically in newly diagnosed disease with the stated chemotherapy backbone or standard-of-care chemotherapy combination.
- Evidence comparing differentiated efficacy, therapeutic index, and tolerability of KRAS versus EGFR inhibition in the stated setting.
- The specific KRAS allele or EGFR activating alteration present in the proposed disease model; both hypotheses are strongly alteration-dependent.
| |
| KRAS — ALK | Human natural experiments | B wins | gpt-5.6-luna | 96% | ALK is better supported by human pharmacological natural experiments: randomized clinical studies in ALK-positive NSCLC show superior efficacy versus chemotherapy and durable benefit with ALK inhibition (C2-EV_CLINICAL_0001, C2-EV_CLINICAL_0002, C2-EV_CLINICAL_0004). Human regulatory safety evidence describes a generally manageable profile while also documenting important toxicities (C2-EV_SAFETY_0005), so tolerability is supported but not unqualified. KRAS has human tumor-genomic driver evidence and germline RASopathy evidence, but the supplied germline variants are activating and associated with developmental disease, not human loss-of-function or protective phenotypes supporting inhibition; the required genetic-direction match is therefore absent (C1-CAT-ONCO-BIO-005, C1-CAT-ONCO-BIO-016, C1-CAT-ONCO-BIO-056). | - For KRAS, a human loss-of-function or protective-allele study, or a long-term human KRAS-inhibitor exposure cohort directly reporting both NSCLC efficacy and tolerability, is missing. The supplied KRAS germline disease evidence reflects activating variants and therefore does not match therapeutic inhibition.
| - ALK is clinically validated as a predictive therapeutic biomarker in advanced NSCLC by randomized phase 3 evidence showing superior efficacy of ALK inhibition over platinum-pemetrexed chemotherapy.
- Second-generation ALK inhibition provides superior systemic and intracranial disease control compared with first-generation therapy in untreated advanced ALK-positive NSCLC.
- Lorlatinib provides durable long-term clinical benefit in treatment-naive advanced ALK-positive NSCLC, with sustained systemic and intracranial disease control.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- The pivotal PROFILE 1014 trial did not demonstrate a statistically significant overall-survival benefit for first-line crizotinib in the unadjusted analysis.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
- Gibt es genetische Erkrankungen, die mit KRAS in Verbindung stehen?
- Gibt es Erkrankungen, die mit genetischen Keimbahnveränderungen von KRAS assoziiert sind?
- Ist KRAS in der Keimbahn genetisch verändert?
|
| ALK — EGFR | Combination potential | Abstained | gpt-5.6-luna | 98% | Neither dossier provides direct combination preclinical evidence with chemotherapy or an overlapping-toxicity/co-dosing assessment, both of which are central to Q12 and required by the guardrail. The dossiers contain target-specific resistance data, but it concerns resistance to ALK or EGFR inhibitors rather than combination with chemotherapy. Reported single-agent toxicities for ALK and EGFR indicate potential safety considerations but do not establish whether either combination can be dosed safely. Therefore, the evidence is insufficient to compare combination potential. | - Direct ALK-plus-chemotherapy and EGFR-plus-chemotherapy combination studies in relevant NSCLC models, including pathway complementarity and antitumor interaction data.
- Combination-specific overlapping-toxicity and co-dosing assessments for each target with chemotherapy.
- Chemotherapy-relevant resistance-mechanism data showing whether ALK or EGFR inhibition prevents or delays resistance to the chemotherapy backbone.
| |
| KRAS — ALK | Novelty, crowding, and strategic optionality | A wins | gpt-5.6-luna | 67% | KRAS has the broader documented strategic platform: clinically validated mutation-specific targeting with expansion toward G12D, G12V, pan-KRAS/pan-RAS approaches, alternative binding sites, combinations, and relevance across multiple tumor types. ALK has strong platform fit and documented expansion into adjuvant, resectable disease, but its strategic value is explicitly constrained by a highly crowded NSCLC segment with several preferred first-line TKIs and emerging fourth-generation competitors. Crowding raises biological confidence for ALK while reducing differentiation and strategic value. The comparison is narrow because KRAS crowding and patent position are not documented, so its apparent crowding advantage cannot be treated as established. | - A directly comparable competitor-pipeline inventory by phase for KRAS versus ALK in NSCLC and other indications.
- Patent-landscape evidence, including blocking patents, freedom-to-operate, and remaining exclusivity for both targets.
- Direct evidence quantifying KRAS competitive crowding in NSCLC; the ALK dossier documents heavy crowding, but the KRAS dossier does not.
| |
| MET — ALK | Evidence robustness (meta-evidence) | B wins | gpt-5.6-luna | 82% | ALK has the more robust meta-evidence package, narrowly. Its support spans distinct source families and multiple apparently independent randomized clinical studies, while also containing explicit negative or discordant findings on overall survival, diagnostic validity, and nonuniform pathway dependence (C2-EV_CLINICAL_0001 through C2-EV_CLINICAL_0006; C2-EV_BIOMARKER_0006; C2-EV_MECHANISM_0006; C2-EV_MECHANISM_0007). MET also has substantial contradictory and negative evidence, including failed broad-MET trials and biomarker discordance, but several prominent efficacy and validation claims are explicitly linked to shared GEOMETRY, VISION, or LUMINOSITY cohorts, reducing their independence. The dossiers do not provide complete overlap or publication-bias metadata, so the margin is narrow rather than clear. | - Explicit cohort-overlap and source-family metadata for the ALK clinical, genomic, and resistance studies; a systematic assessment of unpublished negative studies or publication bias for both candidates; replication in the specified mucus-producing glandular-cell and patient-derived-xenograft context.
| |
| ALK — KRAS | Human natural experiments | A wins | gpt-5.6-luna | 84% | ALK has direct human pharmacological evidence in NSCLC showing substantial efficacy of inhibition, including randomized comparisons against chemotherapy and durable long-term benefit, together with clinical safety evidence documenting tolerability and adverse-event burden (C1-EV_CLINICAL_0001, C1-EV_CLINICAL_0004, C1-EV_SAFETY_0005, C1-EV_SAFETY_0007). Its human genetic evidence is mainly somatic gain-of-function oncogenic evidence rather than a human inhibition natural experiment, and germline ALK evidence does not establish an NSCLC phenotype (C1-EV_GENETICS_0004). KRAS has strong human somatic gain-of-function evidence and established small-molecule tractability, but the dossier does not supply direct human NSCLC efficacy-and-tolerability outcomes for KRAS inhibition; the safety item is a general mechanistic claim rather than a clinical exposure cohort (C2-EV_SAFETY_0001). Therefore ALK is better supported, but only narrowly under this criterion and with the genetic-direction/time-limited-exposure guardrail noted. | - For KRAS: direct human NSCLC clinical evidence showing efficacy and tolerability of KRAS inhibition, preferably from a prospective treatment cohort or randomized trial. For both targets: human loss-of-function or long-term exposure evidence specifically demonstrating tolerability of therapeutic-direction inhibition, with explicit accounting for the mismatch between lifetime genetic exposure and time-limited pharmacological inhibition.
| - ALK is clinically validated as a predictive therapeutic biomarker in advanced NSCLC by randomized phase 3 evidence showing superior efficacy of ALK inhibition over platinum-pemetrexed chemotherapy.
- Second-generation ALK inhibition provides superior systemic and intracranial disease control compared with first-generation therapy in untreated advanced ALK-positive NSCLC.
- Lorlatinib provides durable long-term clinical benefit in treatment-naive advanced ALK-positive NSCLC, with sustained systemic and intracranial disease control.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
- Authoritative germline ALK susceptibility is established for neuroblastic tumors, not NSCLC.
- A large NSCLC germline sequencing study found that inherited susceptibility explains only a minority of cases and did not establish ALK as an NSCLC predisposition gene.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- Mutant-selective inhibitors spare wild type
|
| KRAS — ALK | Perturbational evidence | Abstained | gpt-5.6-luna | 93% | The dossiers contain perturbational evidence for both targets, but neither provides sufficiently explicit, outcome-level evidence of target modulation reversing disease-relevant phenotypes in a patient-derived xenograft. KRAS has a generic statement that tool compounds were used in NSCLC PDX models (C1-CAT-ONCO-BIO-010), while the cited xenograft result is described as a human xenograft rather than a PDX (C1-CAT-ONCO-BIO-117b). ALK has direct knockdown and CRISPR evidence, including in-vivo resistance validation, but the supplied records do not clearly identify a PDX reversal experiment (C2-EV_PERTURBATION_0001, C2-EV_PERTURBATION_0002). Therefore, the requested PDX-specific comparison cannot be adjudicated. | - For each target, an explicitly identified patient-derived xenograft study showing target perturbation, appropriate controls, and reversal of a disease-relevant tumor phenotype such as tumor regression or durable growth suppression; the KRAS dossier only generically mentions PDX use without reporting a specific PDX outcome, while the ALK dossier reports perturbation and in-vivo validation without clearly establishing a PDX reversal experiment.
| |
| KRAS — ALK | Competitive differentiation | B wins | gpt-5.6-luna | 97% | ALK offers greater potential for clinically meaningful differentiation over chemotherapy because ALK-directed therapy has direct randomized evidence in previously untreated, biomarker-selected NSCLC showing improved progression-free survival and response rate versus platinum-pemetrexed chemotherapy (C2-EV_CLINICAL_0001, C2-EV_BIOMARKER_0002). The dossier also supports clinically meaningful differentiation through intracranial efficacy and durable long-term disease control (C2-EV_TRACTABILITY_0003, C2-EV_COMPETITIVE_0002), with mechanistic differentiation against resistance mutations such as G1202R (C2-EV_COMPETITIVE_0004). Although the ALK segment is highly competitive and differentiation among current TKIs is not uniformly established (C2-EV_COMPETITIVE_0001, C2-EV_COMPETITIVE_0006), the evidence directly addresses the stated chemotherapy comparison. KRAS has strong biological and small-molecule tractability evidence, but the supplied dossier does not provide comparable direct evidence of superiority over chemotherapy in newly diagnosed NSCLC. This judgment concerns scientific and clinical differentiation evidence, not market attractiveness. | - A direct randomized comparison of a KRAS-directed therapy with chemotherapy in newly diagnosed KRAS-altered NSCLC.
- A current, disease-specific KRAS competitor pipeline and treatment-landscape analysis addressing differentiation over first-line chemotherapy.
| |
| MET — EGFR | Time to decisive experiment | Abstained | gpt-5.6-luna | 99% | The dossiers do not provide the organization-specific evidence required to compare time or cost to a high-information in vivo proof-of-concept decision. MET has evidence of prior MET-amplified lung-cancer PDX work and validated pharmacodynamic assays (C1-EV_PERTURBATION_0002, C1-EV_BIOMARKER_0005), while EGFR has general evidence that tool compounds and preclinical models exist (C2-CAT-ONCO-BIO-010, C2-CAT-ONCO-BIO-115). However, neither dossier supplies internal capability, model availability for the specified context, capacity, historical cycle times, or costs. These omissions prevent a defensible ranking; understudied or nonspecific evidence cannot establish that either candidate is faster or cheaper. | - Organization-specific assay and reagent inventory for MET and EGFR, including procurement or access times.
- Organization-specific readiness, throughput, and validation status of the required lung NSCLC mucus-producing glandular-cell and patient-derived xenograft models.
- Internal personnel, facility, and operational capacity for the in vivo proof-of-concept studies.
- Historical organizational cycle times and costs from target selection through interpretable in vivo proof-of-concept results for both targets.
| – |
| ALK — MET | Patient stratification | A wins | gpt-5.6-luna | 84% | Both targets support prospective, operational patient selection. ALK is clearer because a single actionable molecular class—ALK rearrangement—is prospectively testable with FDA-authorized assays and was used for selection in randomized clinical evidence demonstrating benefit over chemotherapy (C1-EV_BIOMARKER_0001, C1-EV_BIOMARKER_0002, C1-EV_TRACTABILITY_0005, C1-EV_TRACTABILITY_0006, C1-EV_CLINICAL_0001). METex14 is also a strong prospective hypothesis, supported by tissue RNA or plasma NGS selection and regulatory indication (C2-EV_BIOMARKER_0001, C2-EV_CLINICAL_0003), but MET amplification and protein-based alternatives have substantial threshold dependence, assay discordance, and heterogeneity (C2-EV_BIOMARKER_0006, C2-EV_TISSUE_0004, C2-EV_CLINICAL_0006). The margin is narrow because METex14 itself is clinically well operationalized. | - A direct prospective comparison of standardized ALK-rearrangement and METex14 genomic-testing workflows in the same NSCLC trial population, including assay failure and discordance rates.
| |
| KRAS — MET | Human genetics and causal evidence | Abstained | gpt-5.6-luna | 95% | Both dossiers provide recurrent somatic tumor-sequencing evidence for direct alterations in the named genes, including KRAS activating mutations and MET exon-14 splice alterations, but they do not provide the specified human-genetic causal-evidence types needed for this comparison: GWAS/fine-mapping, L2G assignment, molecular-QTL colocalisation, rare-variant or biobank evidence, or ancestry-diverse replication. Somatic recurrence alone does not establish the requested comparative human-genetic causal evidence, so ranking the candidates would overinterpret incomplete dossiers. | - GWAS or fine-mapping results linking NSCLC risk loci to KRAS or MET
- L2G-style variant-to-gene assignment scores
- Molecular QTL colocalisation in lung or mucus-producing glandular cells
- Rare-variant burden, Mendelian randomization, or biobank phenotype evidence for NSCLC causality
- Ancestry-stratified replication and effect estimates
| – |
| MET — KRAS | Human natural experiments | A wins | gpt-5.6-luna | 78% | MET is better supported because human pharmacological inhibition studies in MET exon 14-skipping NSCLC show substantial response rates in both treatment-naive and previously treated patients, with clinically characterized but monitorable toxicity (C1-EV_SAFETY_0002, C1-EV_SAFETY_0005, C1-EV_CLINICAL_0001, C1-EV_CLINICAL_0002). The conclusion is narrow because these are clinical treatment studies rather than clear lifetime natural experiments, and the MET genetic evidence describes activating somatic alterations; under the guardrail, that genetic direction should not be treated as equivalent to human loss-of-function support for inhibition. KRAS has human activating-driver evidence and a general statement that mutant-selective inhibitors spare wild-type KRAS (C2-EV_SAFETY_0001), but the dossier supplies no direct NSCLC human efficacy-and-tolerability evidence for KRAS inhibition. The available KRAS germline evidence concerns activating RASopathies and therefore does not provide protective loss-of-function support for inhibition. | - For MET, a human biobank loss-of-function or protective-allele study, or a long-term real-world MET-inhibitor exposure cohort, would more directly satisfy the natural-experiment criterion.
- For KRAS, direct human clinical evidence in NSCLC showing efficacy and tolerability of KRAS inhibition is missing from the supplied evidence.
- For both candidates, the supplied human genetic evidence is primarily activating somatic disease biology rather than protective human loss-of-function evidence; this does not directly establish tolerability of therapeutic inhibition.
| |
| EGFR — MET | Cell and tissue relevance | Abstained | gpt-5.6-luna | 96% | The dossiers provide disease-relevant evidence in broad NSCLC tumor-cell populations, particularly MET immunohistochemistry and its substantial discordance with MET genomic alterations. EGFR evidence is also broad and describes expression in malignant epithelial cells. However, neither dossier specifically evaluates the designated mucus-producing glandular cells using single-cell/spatial methods or cell-resolved pathology, nor does either control for composition shifts and disease stage. Because the critical cell type is not directly characterized for either target, the candidates cannot be validly ranked under this criterion. | - Single-cell or spatial transcriptomic, proteomic, or immunohistochemical evidence specifically resolving EGFR and MET expression/localisation/dysregulation within mucus-producing glandular cells in human NSCLC, with control for cell-composition shifts and disease stage.
| |
| EGFR — ALK | Time to decisive experiment | Tie | gpt-5.6-luna | 76% | Both targets have readily available small-molecule tools and established assay/model precedents relevant to an in vivo proof-of-concept. EGFR has multiple tool compounds and published xenograft, PDX, and mouse-model validation precedents; ALK has several approved TKIs, established companion-diagnostic infrastructure, direct PDX perturbation evidence, and genetically engineered lung-tumor models. The dossiers do not provide organisation-specific capacity, historical cycle times, or cost data, so neither target is supported as faster or cheaper. On the available readiness evidence, the candidates are comparable. | - Internal assay and in vivo model capability/capacity for this organisation
- Historical cycle times from target selection through decisive PDX proof-of-concept
- Comparative cost and turnaround data for EGFR versus ALK assays, reagents, biomarker testing, and PDX studies
- Readiness of a PDX model specifically representing mucus-producing glandular cells in NSCLC
| |
| EGFR — KRAS | Integrated therapeutic hypothesis | B wins | gpt-5.6-luna | 72% | Both candidates have coherent genetically defined NSCLC driver hypotheses, tractable small-molecule modalities, and supporting mechanistic and preclinical evidence. EGFR has particularly strong causal validation in NSCLC and multiple approved small-molecule inhibitor generations, but the supplied evidence also documents on-target wild-type EGFR dermatologic and gastrointestinal toxicity, and EGFR inhibition is less inherently differentiated because it affects a broadly expressed receptor. KRAS has strong recurrent human NSCLC driver evidence, clinically validated mutant-selective covalent small-molecule tractability, and supplied evidence that wild-type KRAS is relatively spared. Those features better support a differentiated medicine hypothesis. The advantage is narrow because the strongest therapeutic evidence is allele-specific to KRAS G12C, while the candidate description also includes G12D, and the dossiers do not provide a matched study in the specified cell type, newly diagnosed setting, PDX context, or chemotherapy combination. | - Head-to-head, allele-stratified PDX efficacy and combination data in newly diagnosed NSCLC mucus-producing glandular cells with chemotherapy, including direct comparison of KRAS inhibition versus EGFR inhibition.
- Direct evidence that the proposed KRAS modality covers the relevant allele in this context: G12C has clinical tractability, whereas equivalent therapeutic validation for G12D is not supplied.
- Matched therapeutic-window and normal-lung toxicity data for both candidates under the specified chemotherapy backbone.
| |
| EGFR — KRAS | Cell and tissue relevance | Abstained | gpt-5.6-luna | 95% | The dossiers provide generic evidence for EGFR expression or overexpression in malignant epithelial NSCLC cells (C1-CAT-ONCO-BIO-047, C1-CAT-ONCO-BIO-058, C1-CAT-ONCO-BIO-097) and for broad, largely ubiquitous KRAS expression with mutation-driven rather than generally expression-driven dysregulation (C2-CAT-ONCO-BIO-047, C2-CAT-ONCO-BIO-058, C2-CAT-ONCO-BIO-097). However, neither dossier supplies cell-type-resolved evidence in the specified mucus-producing glandular cells, localisation data for that compartment, or composition- and stage-controlled human pathology. These gaps prevent a criterion-specific comparison. | - Cell-type-resolved single-cell or spatial transcriptomics, proteomics, or immunohistochemistry in human NSCLC specifically identifying mucus-producing glandular cells and measuring EGFR versus KRAS expression, localisation, and dysregulation.
- Analyses controlling for cell-composition shifts and disease stage to distinguish target dysregulation from altered abundance of mucus-producing glandular cells.
- Direct human pathology evidence comparing EGFR and KRAS in the specified lung cell type.
| – |
| MET — KRAS | Biomarker and target engagement | A wins | gpt-5.6-luna | 97% | MET offers the more credible biomarker strategy. MET exon 14 skipping has prospective clinical selection using tissue RNA or plasma cfDNA/NGS, an on-treatment ctDNA depletion pharmacodynamic readout associated with radiographic response, and a validated pMET:total-MET assay for direct target engagement. Capmatinib and tepotinib approvals using FDA-approved tests provide strong regulatory precedent. The strategy is not perfect: MET IHC alone failed as a predictive biomarker and intratumoral heterogeneity can impair classification. KRAS has evidence of clinically validated small-molecule tractability, but the supplied dossier does not provide comparable direct NSCLC evidence for a clinical selection assay, target engagement/pharmacodynamics, or linked early efficacy signal; its relevant evidence is primarily generic or catalogue-level. Therefore FIRST is clearly better for this criterion. | - For KRAS: prospective NSCLC clinical evidence linking a defined genomic biomarker to patient selection, a validated clinical target-engagement or pharmacodynamic assay, and an early efficacy signal with regulatory companion-diagnostic precedent.
| |
| ALK — EGFR | Human natural experiments | Abstained | gpt-5.6-luna | 94% | Neither dossier contains the criterion-specific human natural-experiment evidence needed to compare both efficacy and tolerability under inhibition. ALK has somatic gain-of-function evidence and pharmacological clinical efficacy, but no human biobank LoF/protective-allele or long-term exposure cohort; its germline evidence is for neuroblastic tumors rather than NSCLC (C1-EV_GENETICS_0004). EGFR has reported germline predisposition and activating variants (C2-CAT-ONCO-BIO-016; C2-CAT-ONCO-BIO-056), but no documented human natural experiment jointly establishing inhibitory efficacy and tolerability, and the lifetime genetic exposure does not directly match time-limited inhibition. The available safety evidence instead indicates on-target toxicity for EGFR inhibition (C2-EV_SAFETY_0001) and clinically important toxicity for ALK inhibition (C1-EV_SAFETY_0005). Therefore, the dossiers are insufficient for a criterion-specific ranking. | - For ALK: a human biobank loss-of-function or protective-allele study, or a long-term human ALK-inhibitor exposure cohort that jointly evaluates efficacy and tolerability.
- For EGFR: a human gain- or loss-of-function natural experiment with therapeutic-direction concordance, plus human pharmacological exposure evidence jointly demonstrating efficacy and tolerability.
- For both candidates: explicit evidence connecting the inherited or lifetime genetic perturbation to the proposed time-limited pharmacological inhibition; lifetime genetic exposure cannot by itself establish this equivalence.
| |
| MET — ALK | Reversibility and therapeutic window | Tie | gpt-5.6-luna | 67% | Both candidates use oral small-molecule kinase inhibitors, supporting titration and treatment interruption as a modality-level reversibility precedent (C1-EV_TRACTABILITY_0002; C2-EV_TRACTABILITY_0002). MET has a stated manageable on-target profile, but clinically important edema, interstitial lung disease, hepatotoxicity, and treatment discontinuation are documented (C1-EV_SAFETY_0001; C1-EV_SAFETY_0004; C1-EV_SAFETY_0006; C1-EV_SAFETY_0007). ALK has evidence supporting limited adult normal-tissue essentiality, but also substantial hepatotoxic, pulmonary, cardiac, neurologic, and metabolic toxicities (C2-EV_SAFETY_0001; C2-EV_SAFETY_0004; C2-EV_SAFETY_0005; C2-EV_SAFETY_0006; C2-EV_SAFETY_0007). The dossiers therefore support comparable, clinically manageable but nontrivial windows and comparable small-molecule reversibility; they do not provide the exposure-response, half-life, or interruption-offset data needed to distinguish them. | - Direct exposure-response separation for each target
- Modality-specific half-life, pharmacodynamic offset, and duration after dose interruption
- Human genetic dose-phenotype relationships relevant to normal-tissue essentiality
- Clinical evidence quantifying reversibility after interruption or discontinuation of adverse effects
| - Manageable on-target profile
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- MET is clinically tractable in NSCLC through genotype-selected, oral small-molecule kinase inhibition of MET exon 14–skipping tumors.
- Favourable on-target safety
- Wild-type ALK appears to have limited essentiality for survival of adult normal tissues, although loss of ALK produces specific reproductive and neuroendocrine phenotypes in mice.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- ALK is clinically tractable with an established oral small-molecule kinase-inhibitor modality in metastatic NSCLC.
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- ALK inhibition has a generally manageable clinical safety profile but carries clinically important risks, including hepatotoxicity, interstitial lung disease/pneumonitis, QT prolongation, bradycardia, severe visual loss, and embryo-fetal toxicity.
- Alectinib has clinically important hepatotoxicity and other organ toxicities that complicate the safety case for ALK inhibition.
- First-line lorlatinib produces a high burden of metabolic, neurologic and other adverse events, with more grade 3–4 toxicity than crizotinib.
|
| EGFR — MET | Disease biology and mechanism coherence | B wins | gpt-5.6-luna | 86% | Both targets have coherent NSCLC biology. EGFR is supported by causal activating mutations and ligand-independent activation of MAPK/PI3K pathways (C1-EV_GENETICS_0001, C1-EV_MECHANISM_0001), but much of the detailed support is catalogue-level and less directly tied to a defined inhibition experiment. MET has a more tightly connected evidence chain: recurrent human MET exon-14 alterations activate the receptor by removing CBL-mediated down-regulation (C2-EV_GENETICS_0002, C2-EV_GENETICS_0003, C2-EV_MECHANISM_0002), selective genetic dependence in amplified NSCLC (C2-EV_PERTURBATION_0001), pharmacologic suppression in vivo including PDX models (C2-EV_PERTURBATION_0002), and single-agent MET dependence in EGFR-mutant/MET-amplified patient-derived models (C2-EV_PERTURBATION_0003). The MET mechanism is biomarker-restricted rather than universal, and resistance/co-alteration evidence limits generality, so the advantage is narrow rather than clear. | - Direct, matched perturbation studies in the specified mucus-producing glandular lung-cell context for both targets, including inhibition-linked pathway and disease-phenotype readouts in comparable NSCLC PDX models.
| |
| KRAS — EGFR | Druggability and tractability | B wins | gpt-5.6-luna | 78% | Both targets have substantial small-molecule tractability evidence in NSCLC. KRAS has clinically validated mutant-selective covalent inhibition through the switch-II pocket, whereas EGFR has a broader and more mature small-molecule precedent with multiple approved TKI generations, established kinase-pocket targeting, and an accessible membrane receptor architecture. EGFR therefore ranks slightly higher on overall tractability. The margin is narrow because EGFR inhibition has on-target wild-type toxicity, while KRAS offers stronger mutant selectivity; neither dossier supplies direct delivery or target-engagement data in the specified lung glandular-cell context. | - Direct comparative experimental structure and pocket-quality metrics for the relevant NSCLC variants.
- Small-molecule exposure, biodistribution, and target-engagement data specifically in lung mucus-producing glandular cells or lung PDX tumors.
- Protein-turnover and residence-time measurements for both targets in the specified cell type.
- Head-to-head selectivity, therapeutic-window, and delivery data for KRAS versus EGFR in the specified lung context.
| |
| MET — EGFR | Safety and essentiality | Abstained | gpt-5.6-luna | 88% | The dossiers contain safety evidence for both targets, but they are not sufficiently comparable for a favourable on-target safety ranking. MET has a reported manageable profile, yet selective MET-inhibitor labels document clinically important edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity. Its key normal-tissue essentiality and expression-breadth evidence is largely absent, so the lack of additional toxicity evidence cannot be treated as evidence of safety. EGFR has evidence of broad epithelial expression, developmental and epithelial defects after knockout, and known on-target dermatologic and gastrointestinal toxicity. These findings indicate real safety liabilities but do not establish that EGFR is less favourable than MET without the missing MET normal-tissue and essentiality evidence. | - For MET: human LoF data, pan-essentiality scores, normal-tissue expression breadth, and relevant knockout phenotypes are missing; these are necessary to distinguish genuine safety from understudy.
- A directly comparable small-molecule, inhibition-specific class-toxicity comparison between MET and EGFR is missing.
| - Manageable on-target profile
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- On-target dermatologic and GI toxicity
- Welcher Phänotyp tritt auf, wenn EGFR deletiert wird?
- Welche Rolle spielt EGFR in gesunden Zellen/Geweben?
- Welche Gewebe oder Organsysteme sind bei der Maus nach dem Knockout des EGFR-Gens betroffen?
- Clinical targeting of MET has substantial but monitorable toxicity, including edema, interstitial lung disease, hepatotoxicity, pancreatic toxicity, and embryo-fetal toxicity.
- Tepotinib has clinically important, potentially fatal pulmonary toxicity in MET-altered NSCLC.
- Tepotinib safety includes hepatotoxicity, pancreatic enzyme elevations and embryo-fetal toxicity requiring monitoring or treatment modification.
- On-target dermatologic and GI toxicity
|
| KRAS — MET | Disease biology and mechanism coherence | B wins | gpt-5.6-luna | 82% | MET has the more direct and experimentally connected causal chain for inhibition in NSCLC: MET exon 14 skipping removes the CBL-binding juxtamembrane region, reduces receptor down-regulation, sustains phospho-MET and downstream MAPK signaling, and is supported by selective genetic knockdown and pharmacologic responses in MET-amplified NSCLC PDX models (C2-EV_MECHANISM_0002, C2-EV_PERTURBATION_0001, C2-EV_PERTURBATION_0002, C2-EV_PERTURBATION_0003). KRAS has strong human-genetic and pathway support as a recurrent activating NSCLC driver (C1-EV_GENETICS_0002, C1-EV_MECHANISM_0001), but the supplied high-quality experimental perturbation evidence is less directly documented for the stated NSCLC PDX and cell context. The margin is narrow because MET dependence is restricted to molecularly defined subsets and is not general across NSCLC, as shown by non-amplified models and A549 knockout data (C2-EV_PERTURBATION_0004, C2-EV_PERTURBATION_0005), while KRAS is strongly established as a causal oncogenic pathway. This judgment addresses mechanistic coherence and experimental support, not druggability. | - A head-to-head, allele-specific KRAS inhibition or genetic ablation study in NSCLC patient-derived xenografts, ideally in the specified mucus-producing glandular-cell context, compared with matched MET perturbation data.
| |
| KRAS — ALK | Combination potential | Abstained | gpt-5.6-luna | 98% | ABSTAIN. The criterion requires pathway complementarity, resistance-mechanism data, combination preclinical studies, and overlapping-toxicity assessment specifically relevant to combination with chemotherapy. The KRAS dossier provides resistance and combination evidence mainly for SHP2/SOS1 or other pathway-directed combinations, not chemotherapy, while the ALK dossier provides chemotherapy comparator evidence and substantial ALK-inhibitor safety data but no concurrent chemotherapy-combination studies or overlap-toxicity assessment. Mechanistic rationale alone cannot overcome the guardrail that the combination must be safely doseable. | - For both KRAS and ALK: direct chemotherapy-combination preclinical studies in NSCLC models, including efficacy versus monotherapy and evidence of pathway complementarity with chemotherapy.
- For both KRAS and ALK: formal overlapping-toxicity and dose-feasibility assessment for concurrent administration with chemotherapy.
- For KRAS: chemotherapy-specific resistance-mechanism or combination data; the supplied combination evidence concerns other pathway inhibitors rather than chemotherapy (C1-CAT-ONCO-BIO-015, C1-CAT-ONCO-BIO-049b).
- For ALK: chemotherapy-specific combination studies and safety assessment; chemotherapy comparator trials do not establish that concurrent combination is feasible (C2-EV_BIOMARKER_0002, C2-EV_CLINICAL_0001).
| – |