Version history

{
  "session_id": "sess_SDA-2026-04-25-gapdebate-afba1a80bd_c747c608",
  "round_number": 3,
  "agent_persona": "persona-domain_expert",
  "agent_backend": "codex_cli",
  "action": "assess",
  "content": "**Triage**\n\nThe ideas worth carrying forward are `6`, `5`, `1`, `2`, `4`, and `7`, in that order. I would drop `3` for now; it is too speculative to support a drug program.\n\nThe main translational point is that only one of these is close to a druggable thesis today: `block classical-pathway activation downstream of C1q` rather than trying to solve all C1q biology first. The rest are mostly mechanism, stratification, or endpoint-selection hypotheses.\n\n**Per Idea**\n\n`6. Spare C1q recognition, block C1r/C1s activation`\n- **Feasibility:** Highest. This is the cleanest therapeutic hypothesis because the target class already exists. C1s inhibition is clinically validated outside CNS, which materially reduces target-risk.\n- **Druggability:** High for systemic antibodies; medium for CNS disease because BBB exposure is the bottleneck. Best near-term route is a peripherally dosed biologic with demonstrated CSF target engagement, not a brand-new CNS-only chemistry program.\n- **Biomarkers:** `CH50/classical pathway activity`, free/total `C1q`, `C4d/C4a`, `C3 fragments`, CSF `C1q/C3`, synaptic injury markers (`neurogranin`, `SNAP-25`), `NfL`, and ideally `SV2A PET`.\n- **Model systems:** Human iPSC neuron-microglia-astrocyte tri-cultures with human complement; complement-sufficient synaptosome assays; tau plus amyloid mouse models; NHP PK/CSF target-engagement work if moving a biologic.\n- **Safety:** Better than pan-`C3/C5`, but not trivial. Risks are infection susceptibility, impaired immune-complex clearance, and chronic innate-immune perturbation. Vaccination/infection monitoring would be required.\n- **Timeline/cost:** Repurposing-like translational package `18–30 months`, `$15M–35M` to get to a strong go/no-go package; new CNS-enabled program to IND more like `4–6 years`, `$40M–90M`.\n- **Verdict:** Lead program.\n\n`5. TREM2 state determines whether C1q tagging is adaptive or toxic`\n- **Feasibility:** Good biology, but therapeutically more complicated than `6` because TREM2 effects are stage- and context-dependent.\n- **Druggability:** Medium. TREM2 antibodies/agonists and peptide approaches are plausible, but the risk is that boosting microglia in the wrong disease stage worsens inflammatory injury.\n- **Biomarkers:** `sTREM2`, CSF `C1q/C3`, `GFAP`, `NfL`, synaptic markers, plaque/tau PET, and if possible `SV2A PET`. This is a biomarker-heavy program.\n- **Model systems:** Isogenic `TREM2 WT/R47H/KO` human microglia in tri-culture; matched opsonized-substrate assays; tau and amyloid+trem2 mouse models.\n- **Safety:** Microglial overstimulation, edema/inflammatory worsening, pathology-stage mismatch, and uncertain chronic dosing window.\n- **Timeline/cost:** `2–3 years`, `$20M–50M` to de-risk the biology; full clinical path in AD would likely be `6–8+ years`, `$80M+`.\n- **Verdict:** Strong modifier/combination axis, not the first standalone bet.\n\n`1. Synaptic C1q drives pruning; microglial surface C1q signals state`\n- **Feasibility:** Biologically plausible, but still not cleanly orthogonalized from ligand identity and microglial state.\n- **Druggability:** Medium-low as stated. “Location-selective C1q modulation” is not yet a drug format. Practical targets would still be `C1q`, `C1s`, `CR3`, or receptor-blocking decoys.\n- **Biomarkers:** Same complement/synaptic panel as above, plus high-content imaging of synaptic vs microglial C1q localization in human tissue and co-cultures.\n- **Model systems:** Best tested in humanized tri-culture with engineered C1q tethering and complement-defined media. Mouse work alone will not settle it.\n- **Safety:** If wrong, you build a precision therapy on an artificial distinction and end up with broad complement suppression anyway.\n- **Timeline/cost:** `18–24 months`, `$10M–25M` to resolve the mechanism enough for portfolio decisions.\n- **Verdict:** Worth funding as a decision-enabling biology program, not yet as a direct asset thesis.\n\n`2. Binding partner identity matters more than location`\n- **Feasibility:** Plausible, but underspecified. This is a discovery program, not a near-term development thesis.\n- **Druggability:** Low-medium. Ligand- or complex-selective blockade is hard because C1q is promiscuous and disease tissue contains mixed complexes.\n- **Biomarkers:** Complex-specific IP-MS/proximity assays, CSF EV `C1q`, ligand panels (`Aβ`, neuronal pentraxins), and downstream complement activation markers.\n- **Model systems:** Full-factorial human assays varying `ligand`, `location`, complement competence, and receiver cell state. This is where the answer will come from.\n- **Safety:** Off-target blockade of beneficial debris handling is a real risk if complex selectivity is imperfect.\n- **Timeline/cost:** `2–4 years`, `$20M–45M` before you know whether there is a tractable complex-selective drug concept.\n- **Verdict:** Good platform/discovery work; weak standalone therapeutic hypothesis today.\n\n`4. APOE isoform reshapes the C1q binding landscape`\n- **Feasibility:** Best viewed as a stratification hypothesis. ApoE biology is real; the C1q-centric causal chain is still loose.\n- **Druggability:** Medium only if you target `APOE` biology broadly or use genotype as an enrichment variable. It is not yet a clean C1q program.\n- **Biomarkers:** `APOE` genotype, CSF `ApoE`, `C1q`, `sTREM2`, `GFAP`, amyloid/tau PET, synaptic markers.\n- **Model systems:** Isogenic `APOE3/APOE4` neuron-astrocyte-microglia cultures; plaque-like aggregate systems; human brain proteomics.\n- **Safety:** ApoE interventions are pleiotropic and can affect lipid handling, vascular biology, and amyloid dynamics.\n- **Timeline/cost:** `2–3 years`, `$15M–35M` to establish whether ApoE status meaningfully predicts response to a complement intervention.\n- **Verdict:** Valuable for patient selection and combination design, not as the primary mechanism.\n\n`7. Distinct inhibitory vs excitatory synapse roles`\n- **Feasibility:** Biologically interesting, but therapeutically premature.\n- **Druggability:** Low. This is more likely to improve indication choice, circuit endpoints, and dose window than to define a drug target.\n- **Biomarkers:** `SV2A PET` for global synaptic density, EEG/MEG network hyperexcitability, region-specific imaging, synaptosome proteomics. There is no mature fluid biomarker for inhibitory-vs-excitatory loss.\n- **Model systems:** Circuit-aware slice systems, electrophysiology-heavy co-cultures, region-specific in vivo models.\n- **Safety:** Biggest risk is perturbing network balance without knowing which synapse class you are sparing or harming.\n- **Timeline/cost:** `2–4 years`, `$10M–30M`.\n- **Verdict:** Keep as an endpoint-design subproject, not a lead drug hypothesis.\n\n**Program Recommendation**\n\nIf this were a portfolio review, I would do:\n\n1. Advance `6` as the lead translational hypothesis.\n2. Run `5` in parallel as a response-modifier/combination hypothesis.\n3. Fund `1` as the key mechanism-resolution experiment.\n4. Keep `2`, `4`, and `7` as exploratory stratification packages.\n5. Stop `3` unless direct endogenous `C1q-THBS` complexes are first shown in human brain.\n\nThe most realistic path is not “find the one true role of C1q first.” It is: prove that selective classical-pathway blockade preserves synapses in a humanized system, then use `TREM2`, `APOE`, and circuit-bias biology to decide who should get it and when.\n\nSources: [PMID 18083105](https://pubmed.ncbi.nlm.nih.gov/18083105/), [PMID 27033548](https://pmc.ncbi.nlm.nih.gov/articles/PMC5094372/), [PMID 37442133](https://pubmed.ncbi.nlm.nih.gov/37442133/), [PMID 33628204](https://pubmed.ncbi.nlm.nih.gov/33628204/), [PMID 30692699](https://pubmed.ncbi.nlm.nih.gov/30692699/), [PMID 37118504](https://pubmed.ncbi.nlm.nih.gov/37118504/), [PMID 8176223](https://pubmed.ncbi.nlm.nih.gov/8176223/), [PMID 29202623](https://pubmed.ncbi.nlm.nih.gov/29202623/), [Annexon ANX005 Phase 3 update, June 25 2024](https://ir.annexonbio.com/news-releases/news-release-details/annexon-presented-additional-positive-phase-3-results-anx005-c1q), [PMID 37246953](https://pubmed.ncbi.nlm.nih.gov/37246953/), [PMID 37023079](https://pubmed.ncbi.nlm.nih.gov/37023079/), [PMID 40091552](https://pubmed.ncbi.nlm.nih.gov/40091552/), [SV2A PET in AD](https://pmc.ncbi.nlm.nih.gov/articles/PMC7383876/).\n\nIf you want, I can convert this into a structured scorecard with `druggability / biomarker readiness / model readiness / safety / timeline / cost` scored `1–5` for direct insertion into SciDEX.",
  "tokens_used": "2117",
  "persona_id": "persona-domain_expert"
}