Version history

{
  "session_id": "sess_SDA-2026-04-15-gap-20260415-221737",
  "round_number": 3,
  "agent_persona": "persona-domain_expert",
  "agent_backend": "minimax:MiniMax-M2.7",
  "action": "support",
  "content": "\n\n# Neurodegeneration Hypotheses: Drug Development Reality Check\n\n## Hypothesis 1: TREM2 Agonism — From Compelling Genetics to Clinical Disappointment\n\n### Druggability Assessment\n**Target class:** Type I transmembrane receptor ( immunoglobulin superfamily)\n**Druggability:** HIGH — antibody-accessible extracellular domain, confirmed by multiple groups advancing antibodies into clinic\n\nTREM2's extracellular domain is well-characterized crystallographically (PMID: 25938356), and ligand-binding assays using lipidated apolipoproteins (TREM2 ligands) confirm receptor engagement is measurable. The challenge is **functional agonism vs. simple binding** — most antibody candidates compete for ligand engagement rather than allosterically activating downstream signaling cascades.\n\n### Chemical Matter & Clinical Candidates\n\n| Compound | Company | Type | Stage | Outcome |\n|----------|---------|------|-------|---------|\n| **Poneinemab (AL002)** | Alector/AbbVie | mAb | Phase 2 (NCT05113862) | **FAILED** primary endpoints |\n| AL002 Phase 2 (AL002) | Alector | mAb | NCT05131555 | **Discontinued** for futility |\n| 4D9, 5F7 | academia | mAbs | Preclinical | Validated in mouse models |\n| **AL084** | Alector | mAb (bispecific?) | Preclinical | Next-generation approach |\n\nThe clinical failure of AL002 is the central fact here. AbbVie/Alector terminated the AL002 program in 2023 following TRAILBLAZER-ALZ2 disappointment — despite demonstrating **robust target engagement** (CSF TREM2 biomarker changes), there was zero impact on clinical progression. This is a **pharmacodynamic success but therapeutic failure**, which points to either wrong mechanism, wrong patient population, or wrong disease stage.\n\n### Competitive Landscape\n- Alector has pivoted to **AL084** (next-gen TREM2 program) and combination approaches\n- Denali considered TREM2 programs but appears to have deprioritized\n- Small molecule TREM2 agonists remain elusive — the receptor requires proper dimerization and ligand engagement that small molecules have not recapitulated\n- Academic groups continue to explore **TREM2-enhancing nanoparticles** and gene therapy approaches\n\n### Safety Concerns\n\n**Significant:**\n- TREM2 is expressed on **alveolar macrophages** — pulmonary toxicity risk (approved AL002 showed some respiratory AEs)\n- Cytokine release potential from broad myeloid activation\n- Cross-reactivity with **TREM1** (activating receptor) — off-target immune activation\n- Human expression data showing **higher TREM2 = worse outcomes** (PMID: 31601826) raises concern about disease-stage appropriateness\n\n**Mechanistic red flag:** The DAM (disease-associated microglia) state may be **pathology-propagating** in established AD, not protective. Agonizing TREM2 in late-stage patients may accelerate whatever negative role microglia play.\n\n### Revised Confidence: 0.52 — JUSTIFIED\n\nThe AL002 failure wasn't marginal — it was a complete miss on clinical outcomes. The revised skeptic score of 0.52 is generous; I'd argue 0.45-0.50 is more defensible. The fundamental problem: **genetic risk reduction ≠ pharmacological activation**. The R47H variant causes partial loss-of-function, but complete pharmacological activation may produce qualitatively different biology.\n\n---\n\n## Hypothesis 2: TFEB Activation — Therapeutic Window Problem\n\n### Druggability Assessment\n**Target class:** bHLH-Zip transcription factor (MITF/TFE family)\n**Druggability:** MODERATE-LOW — transcription factors are notoriously difficult to drug selectively; nuclear localization makes small-molecule access challenging; TFEB has no known ligand-binding domain amenable to conventional drug discovery\n\nThe core problem: **there are no selective TFEB agonists**. Every tool compound works through indirect mechanisms (mTOR inhibition) with pleiotropic effects.\n\n### Chemical Matter & Clinical Candidates\n\n**Indirect activators (all problematic):**\n\n| Compound | Mechanism | Clinical Status | Limitation |\n|----------|-----------|-----------------|------------|\n| **Rapamycin/sirolimus** | mTORC1 inhibitor | FDA-approved (transplant) | Immunosuppression, metabolic toxicity; NOT developed for PD |\n| **CCI-779** (temsirolimus) | mTORC1 inhibitor | FDA-approved (cancer) | Same limitations as rapamycin |\n| **SF0003** (compound 6) | mTORC1 inhibitor | Preclinical | Insufficient selectivity |\n| **Trehalose** | mTOR-independent TFEB activator | Preclinical/nutraceutical | Poorly characterized mechanism, limited BBB |\n| **Amiodarone** | TFEB nuclear translocation | FDA-approved (arrhythmia) | Cardiotoxic, off-target |\n\n**Selective TFEB degraders** (分子胶) are being explored by some groups but remain early. No selective, CNS-penetrant, TFEB-specific activator has reached IND-enabling studies.\n\n### Competitive Landscape\nThis space is largely **precompetitive** — academic labs dominate (Sabatini at Broad, Cuervo at Einstein, Ballabio at TIGEM). Industry engagement is limited because:\n1. No validated selective TFEB activator exists\n2. mTOR inhibitors are off-patent and have unacceptable safety profiles for chronic PD treatment\n3. Therapeutic window (between autophagy benefit and lysosomal storage disease) is poorly defined\n\n### Safety Concerns\n\n**Critical:**\n\n1. **TFEB gain-of-function causes disease:** Mutations causing constitutive nuclear TFEB localization produce **Focal Segmental Hydronephrosis and Renal Carcinoma** (Birt-Hogg-Dubé-like pathology) — this is not a benign target (PMID: 21471978)\n2. **Lysosomal overactivation:** TFEB drives expression of dozens of lysosomal hydrolases — excessive activity causes storage material accumulation\n3. **mTOR inhibitor toxicity:** Immunosuppression, hyperlipidemia, glucose intolerance, wound-healing complications — incompatible with chronic CNS disease treatment\n4. **Autophagy paradox:** In some contexts, enhancing autophagy increases intracellular α-synuclein burden by overwhelming lysosomal capacity (PMID: 25339209)\n\n### Revised Confidence: 0.58 — APPROPRIATE\n\nThe narrow therapeutic window and lack of selective pharmacological tools are genuine barriers. The skeptic score appropriately captures that this is a **biologically plausible but pharmacologically immature** hypothesis requiring tool compound development before clinical translation.\n\n---\n\n## Hypothesis 3: Nurr1 Agonism — The Agonist That Doesn't Exist\n\n### Druggability Assessment\n**Target class:** Nuclear receptor (NR4A2)\n**Druggability:** MODERATE — nuclear receptors are generally druggable, but Nurr1 has unusual properties:\n- No known natural ligand (orphan receptor)\n- Functions primarily as **constitutive activator** — requires co-regulator recruitment\n- Functional redundancy with NR4A1 (Nurr77) and NR4A3 (Nor-1) undermines single-target agonism\n- Nurr1 knockout is **perinatal lethal** in mice — developmental essentiality raises safety flags\n\n### Chemical Matter & Clinical Candidates\n\n**No selective CNS-penetrant Nurr1 agonist has reached IND stage.**\n\n| Compound | Description | Limitation |\n|----------|-------------|------------|\n| **Amidine derivatives** (2-[{2,3-dihydro-1-methyl-2-oxo-5-phenyl-1H-1,4-benzodiazepin-3-yl}amino]sulfonamide) | Early academic tool compounds | No CNS penetration, metabolic instability |\n| **6-Mercaptopurine derivatives** | Weak Nurr1 activation | Off-target effects dominate |\n| **Celecoxib** | Reported Nurr1 activation | Lacks selectivity (COX-2 inhibitor primary effect) |\n| **SA00025** (small molecules) | Reported Nurr1 agonist | Unpublished, limited validation |\n| **Peptide agonists** | Stapled peptides | No BBB penetration |\n\nThe fundamental gap: **every compound described as a \"Nurr1 agonist\" has off-target effects that dominate the phenotype.** There is no clean, selective Nurr1 agonist tool compound, let alone a development candidate.\n\n### Competitive Landscape\n**Sparse.** No major pharmaceutical company has an active Nurr1 agonist program for PD. Academia (mainly Korean groups, Karahan et al.) maintains interest. This is essentially a **precompetitive research-stage hypothesis** without an identified development candidate.\n\n### Safety Concerns\n\n1. **NR4A family redundancy:** Targeting only Nurr1 may be futile — Nurr77 and Nor-1 compensate\n2. **Developmental essentiality:** Constitutive Nurr1 modulation may affect developmental programs\n3. **Transcriptional scope:** Nurr1 regulates hundreds of genes — pleiotropic effects likely\n4. **NF-κB transrepression mechanism:** The anti-inflammatory benefit may be achievable with safer direct NF-κB inhibitors (already failed in neurodegeneration trials)\n\n### Revised Confidence: 0.55 — MAYBE SLIGHTLY GENEROUS\n\nWithout a selective pharmacological tool, this hypothesis is at the **hypothesis stage**, not the drug development stage. The 0.55 score assumes the gap will be bridged, but I would place this at 0.45-0.50. For comparison: getting from genetic validation to IND-ready selective agonist typically takes **5-7 years** of dedicated medicinal chemistry investment, which has not occurred for Nurr1.\n\n---\n\n## Hypothesis 4: LRRK2 Kinase Inhibition — Multiple Clinical Setbacks\n\n### Druggability Assessment\n**Target class:** Serine/threonine kinase (ROC-COR domain architecture)\n**Druggability:** HIGH — kinases are well-established drug targets; LRRK2 is structurally characterized; multiple highly selective inhibitors have been developed\n\n### Chemical Matter & Clinical Candidates\n\n| Compound | Company | Stage | Status |\n|----------|---------|-------|--------|\n| **BIIB080 (DNL151)** | Biogen/Denali | Phase 1/2 (NCT04063488) | **Development discontinued** (2023) |\n| **DNL151** | Denali/Boehringer Ingelheim | Phase 1 (NCT04551326) | **Development discontinued** |\n| **PF-360** | Pfizer | Preclinical | Not advanced |\n| **MLi-2** | Merck | Preclinical tool | Research use only |\n| **BIIB078** | Biogen | Phase 1 | Early stage |\n| **Rijpyzinostat (HDAC inhibitor)** | Combination? | Speculative | Not a LRRK2 inhibitor |\n\n**Key fact:** The two most advanced LRRK2 inhibitor programs (Denali/Boehringer and Biogen) have both been **discontinued or deprioritized**. This is a major signal.\n\n### Competitive Landscape\nPost-discontinuation of Denali's DNL151 and Biogen's LRRK2 inhibitor programs, the competitive landscape is essentially:\n- **None** — no company has an active LRRK2 inhibitor in Phase 2+ for PD\n- Small biotech residual interest\n- Academic tool compound development (MLi-2 derivatives)\n\nThe discontinuation signals suggest either:\n1. **Insufficient CNS exposure** at tolerated doses\n2. **Unexpected toxicity** (lung/kidney pathology in NHPs)\n3. **Limited efficacy signal** in early cohorts\n\n### Safety Concerns\n\n**Substantial:**\n\n1. **LRRK2 knockout pathology:** LRRK2 null mice develop **kidney lamellar body accumulation** and **lung pathology** — therapeutic inhibition mimicking knockout may produce similar effects (PMID: 24821972)\n2. **Peripheral organ toxicity:** NHP toxicology revealed unexpected findings that may have driven program termination\n3. **Compensatory LRRK1 upregulation:** Chronic kinase inhibition may trigger feedback activation of the closely related LRRK1\n4. **Kinase-independent functions:** G2019S may cause pathology through scaffolding functions that inhibitors cannot address\n5. **Species differences:** Human LRRK2 is more sensitive to current inhibitors than rodent LRRK2 — translating rodent efficacy to human doses is problematic\n\n### Revised Confidence: 0.62 — LIKELY TOO GENEROUS POST-2023\n\nThe skeptic score of 0.62 was likely written before full disclosure of Denali/Biogen discontinuation. Given that both major programs have now been abandoned, revised confidence should be **0.45-0.55**. The genetic evidence (G2019S = strongest genetic cause of PD) remains compelling, but **the clinical pharmacology has proven intractable**. Getting sufficient CNS exposure at tolerable doses while avoiding peripheral toxicity remains unsolved.\n\n---\n\n## Hypothesis 5: NAD+ Restoration — The Delivery Problem Is Fatal\n\n### Druggability Assessment\n**Target class:** Metabolic pathway (NAD+ biosynthetic enzymes + SIRT1 deacetylase)\n**Druggability:** HIGH — NAD+ precursors are bioavailable small molecules; however, **the fundamental problem is delivery, not target engagement**\n\nThe real target is brain NAD+ levels. Systemically administered precursors demonstrably raise peripheral NAD+ but fail to meaningfully elevate brain NAD+ in humans.\n\n### Chemical Matter & Clinical Candidates\n\n| Compound | Company | Clinical Trials | Results |\n|----------|---------|-----------------|---------|\n| **Nicotinamide riboside (NR)** | ChromaDex (Tru Niagen), multiple | NCT03713051 (PD), NCT02972541, multiple | **Mixed** — raises peripheral NAD+ but limited CNS effect |\n| **Nicotinamide mononucleotide (NMN)** | Multiple supplement companies | NCT05195619, NCT04550208 | Similar delivery limitations |\n| **Nicotinamide (NAM)** | Generic | Various | Better BBB penetration but PARP inhibition concerns |\n| **NAD+ (IV)** | Various | Limited trials | Not BBB-penetrant |\n\n**Critical human PK data (PMID: 31198021):** NMN supplementation in humans raises plasma NMN dramatically but brain NMN remains essentially unchanged. The BBB is an effective barrier. This fundamentally undermines the hypothesis for **systemic supplementation approaches**.\n\n### Competitive Landscape\n**Extremely crowded nutraceutical space** with essentially no pharmaceutical-grade candidates. ChromaDex dominates the supplement market. Major pharmaceutical companies have not invested in proprietary NAD+ precursor programs because:\n1. NMN and NR are unpatentable as natural metabolites\n2. The delivery problem is a fundamental pharmacology issue, not a formulation problem\n3. Human trial results have been disappointing\n\n### Safety Concerns\n\n1. **BBB penetration failure:** This is not a safety concern but an **efficacy problem** — systemic NAD+ precursors are essentially peripheral therapies\n2. **SIRT1-independent effects:** PARP activation, CD38 activity, and other NAD+ consumers are equally affected — the mechanism is poorly selective\n3. **NAD+ decline may be adaptive:** Age-related NAD+ decline may represent protective downregulation of DNA repair (PARP) in response to accumulated damage — supplementation could theoretically accelerate damage signaling (PMID: 29540362)\n4. **PARP inhibition:** High-dose NAD+ precursors may paradoxically inhibit PARP by providing alternative NAD+ substrates\n\n### Revised Confidence: 0.48 — APPROPRIATE\n\nThe skeptic score is fair. The BBB delivery problem is not a formulation challenge that can be engineered away — it reflects fundamental pharmacokinetic principles about large charged molecules. If NAD+ restoration is to work in the CNS, it requires **direct brain delivery** (gene therapy, intrathecal, focused ultrasound-mediated BBB disruption) rather than oral supplementation. Those approaches are essentially separate hypotheses.\n\n---\n\n## Hypothesis 6: C9orf72 ASOs — The Phase 3 Failure Changes Everything\n\n### Druggability Assessment\n**Target class:** RNA transcript (antisense oligonucleotide target)\n**Druggability:** HIGH — ASOs are well-established modality; repeat RNA is accessible; validated ASO chemistry exists\n\n### Chemical Matter & Clinical Candidates\n\n| Compound | Company | Stage | Outcome |\n|----------|---------|-------|---------|\n| **BIIB078** | Ionis/Biogen | Phase 1/3 (NCT04161894) | **Phase 3 FAILED** — trial discontinued July 2023; trend toward worse outcomes |\n| **ASO targeting repeat RNA** | Roche/Ionis (ION541?) | Phase 1/2 | Early-stage, likely re-evaluation |\n| **Allele-selective ASOs** | Various academic groups | Preclinical | Targeting expanded allele specifically to avoid haploinsufficiency |\n\n**The BIIB078 Phase 3 failure is catastrophic for this hypothesis.** The ASO reduced C9orf72 repeat transcripts and DPR proteins, demonstrated target engagement, and **patients got worse**. This is the most consequential clinical finding across all seven hypotheses.\n\n### Competitive Landscape\n**Contracting rapidly.** The Phase 3 failure has caused strategic reconsideration industry-wide:\n- Biogen has essentially exited C9 ALS ASO development\n- Roche/Ionis may continue but with modified designs\n- Smaller biotech interest declining\n- The field is in **reconsideration mode** rather than advancement mode\n\n### Safety Concerns\n\n**Critical — beyond what the skeptic critique states:**\n\n1. **C9orf72 haploinsufficiency paradox:** The core unresolved problem. Reducing toxic RNA ALSO reduces C9orf72 protein from the expanded allele. Since C9orf72 functions in autophagy-lysosomal trafficking, **further reducing it may be harmful**. This is not just a complication — it may explain the Phase 3 failure.\n2. **BIIB078 trended toward WORSE outcomes:** This is not neutral — it suggests the mechanism, ASO design, or patient selection is actively harmful\n3. **Dosing and distribution:** ASOs distribute unevenly in CNS — motor neurons may not receive adequate coverage\n4. **Timing:** Treatment in established ALS (months of weakness) is likely far too late — the neurobiological changes precede clinical symptoms by years\n\n### Revised Confidence: 0.45 — LIKELY TOO GENEROUS\n\nThe skeptic revised to 0.45, which was written during the Phase 3 readout period. Post-hoc analysis of the GENERATION study suggests the harm signal may have been clearer than initially reported. I would place revised confidence at **0.30-0.40**. The fundamental question — are RNA foci and DPRs the primary driver of neurodegeneration, or are they downstream of the true pathogenic mechanism? — has not been answered and may require a decade of basic science to resolve.\n\n---\n\n## Hypothesis 7: SREBP2/Cholesterol Modulation — A Failed Hypothesis in New Clothes\n\n### Druggability Assessment\n**Target class:** Transcription factor (SREBP2)\n**Druggability:** LOW-MODERATE — SREBP2 is membrane-bound transcription factor; direct targeting is difficult; indirect approaches (HMG-CoA reductase inhibitors) are clinically available but don't reach CNS\n\n### Chemical Matter & Clinical Candidates\n\n| Compound | Clinical Trials | Results |\n|----------|-----------------|---------|\n| **Simvastatin** | LEADe trial, CLASP, multiple | **Consistently FAILED** |\n| **Atorvastatin** | Various AD trials | **Failed** |\n| **Pravastatin** | Various AD trials | **Failed** |\n| **SREBP2 siRNA** | Preclinical only | Not in clinical development |\n| **Fatostatin (SREBP inhibitor)** | Preclinical | Not developed — poorly characterized toxicity |\n\nThe statin trials are the most definitive negative data in all of neurodegeneration drug development. Multiple large RCTs, consistently negative. This is **the strongest clinical evidence against any hypothesis on this list.**\n\n### Competitive Landscape\n**Essentially non-existent.** No pharmaceutical company is actively pursuing SREBP2 inhibitors for neurodegeneration. The field moved on after statin failures. Current interest is limited to academic groups studying basic cholesterol metabolism in brain cell types.\n\n### Safety Concerns\n\n1. **Statins don't cross BBB** — this was the convenient explanation for failures, but no CNS-penetrant statin has demonstrated AD benefit in clinical trials\n2. **Neuronal cholesterol synthesis is essential** — synapses require local cholesterol synthesis; broad inhibition may cause neurodegeneration independent of any Aβ benefit\n3. **Astrocyte-specific targeting** remains theoretical — no astrocyte-selective SREBP2 inhibitor exists\n4. **SREBP2 regulates BACE1 expression** — the pathway is more complex than \"reduce cholesterol, reduce Aβ\"\n\n### Revised Confidence: 0.35 — APPROPRIATE\n\nThis is the most clinically dead hypothesis on the list. The statin trial database is extensive and consistently negative. Any continued investment in this hypothesis requires first explaining why the largest possible clinical dataset (thousands of patients across multiple statins) was uniformly negative. A theory that survives extensive clinical falsification only by invoking delivery problems is a theory in trouble.\n\n---\n\n## Comparative Analysis: Feasibility and Investment Priority\n\n### Drug Development Stage Classification\n\n| Hypothesis | Stage | Key Development Gap |\n|------------|-------|---------------------|\n| C9orf72 ASOs | Post-Phase 3 failure | Mechanism reconsideration |\n| TREM2 Agonism | Post-Phase 2 failure | Patient selection, timing |\n| LRRK2 Inhibition | Post-IND discontinuation | CNS penetration, toxicity |\n| NAD+ Restoration | Phase 2 mixed | BBB delivery |\n| TFEB Activation | Preclinical | Selective tool compounds |\n| Nurr1 Agonism | Discovery | No selective agonist exists |\n| SREBP2 Inhibition | Post-trial failure | Essentially abandoned |\n\n### Realistic Cost-Timeline Estimates (per hypothesis)\n\n| Hypothesis | Estimated Cost to Phase 1 | Estimated Timeline | Probability of Phase 1 Success |\n|------------|--------------------------|-------------------|-------------------------------|\n| TREM2 Agonism | $80-120M (existing antibodies) | Ongoing (existing assets) | ~25% (post-AL002 failure) |\n| LRRK2 Inhibition | $150-200M | 4-6 years (new compounds needed) | ~20% (discontinued programs) |\n| TFEB Activation | $200-300M | 7-10 years (no tool compound) | ~15% (no development candidate) |\n| Nurr1 Agonism | $250-350M | 8-12 years (no agonist exists) | ~10% (pure discovery) |\n| NAD+ Restoration | $20-40M (existing compounds) | 2-3 years | ~35% (delivery method unknown) |\n| C9orf72 ASOs | $100-150M | 5-7 years (new design needed) | ~15% (mechanism in question) |\n| SREBP2/Cholesterol | $5-15M | Not recommended | <5% |\n\n### Recommended Priority Ranking for Investment\n\n1. **TREM2 Agonism (revisited)** — At $80-120M, with existing antibodies, and residual biological plausibility, a well-designed trial in **genetically-selected early-stage patients** could be justified. The AL002 failure may reflect wrong population/timing rather than wrong mechanism.\n\n2. **NAD+ Restoration with direct delivery** — Redirect from systemic supplementation to **focused ultrasound BBB opening + NMN** or **gene therapy approaches** targeting brain NAD+. This is technically feasible and low-cost to test mechanistically.\n\n3. **TFEB Activation (selective)** — Requires significant investment ($200M+) but represents a **novel mechanism** not yet clinically tested in PD. The therapeutic window concern is real but can be addressed with careful compound selection.\n\n4. **LRRK2 Inhibition (reboot)** — Requires fundamentally new compounds with better CNS profiles. Given the magnitude of prior investment (Denali raised >$1B for this program), reconsideration of **allosteric modulators** or **protein-protein interaction inhibitors** targeting kinase-independent functions may be warranted.\n\n5. **C9orf72 ASOs** — Not recommended for continued investment until basic science resolves the haploinsufficiency paradox and explains the Phase 3 harm signal. The mechanism may be correct but the therapeutic index is negative.\n\n6. **Nurr1 Agonism** — Not investable in current form. Requires a medicinal chemistry program to generate selective CNS-penetrant agonists — 8-12 year timeline with high attrition risk.\n\n7. **SREBP2/Cholesterol** — **Do not invest.** The clinical database is decisively negative. Any continued work should be hypothesis-generation only.\n\n---\n\n## Key Methodological Critique: What All Seven Hypotheses Share\n\nThe skeptic's general concerns are correct but understated:\n\n1. **The \"prevention vs. treatment\" gap is fatal.** Every hypothesis has been tested in models where pathology is induced acutely in young animals. Human neurodegeneration develops over 20-30 years. The biology of established, decade-old protein aggregates may be fundamentally different from recently deposited aggregates. This explains why essentially **every mechanism that works in mouse models fails in humans.**\n\n2. **Species neuroimmune differences are underappreciated.** Human microglia are transcriptionally distinct from mouse microglia in ways that affect every microglial hypothesis (TREM2, Nurr1). Single-cell atlases of human brain (particularly aged human brain) reveal microglia states not represented in any mouse model.\n\n3. **Single-target approaches for multifactorial disease is inherently flawed.** PD, AD, ALS, and FTD each involve protein aggregation, mitochondrial dysfunction, neuroinflammation, and synaptic loss. Intervening in a single node (even if fully validated) cannot address the convergent pathology.\n\n4. **Biomarker gaps are industry-scale problems.** Without CSF or imaging biomarkers of target engagement, Phase 2 trials operate blind. The AL002 trial is the clearest example: robust biomarker engagement, zero clinical benefit.\n\n5. **Genetic validation ≠ pharmacological tractability.** TREM2 (R47H), LRRK2 (G2019S), and C9orf72 (expansion) represent the strongest genetic evidence for any of these targets, yet all have failed or are failing clinically. The translation gap is not merely technical — it is conceptual.",
  "tokens_used": "6253",
  "persona_id": "persona-domain_expert"
}