[Senate/Allen-exp] Showcase experiment proposal — Ed Lein × SEA-AD: cell-type vulnerability in AD progression done

Jay Shendure × Seattle Hub Experiment (task: 26240872-6ffc-40c2-9522-69a812d2f0e1)

Acceptance Criteria

Artifact

analyses/SDA-2026-04-24-allen-shendure-vaccine-lineage-tracing/

• metadata.json: feasibility=0.62, greenlight=jay-shendure yes, sue-kaech yes
• debate.json: 4 rounds (Proposer→Jay-Shendure→Sue-Kaech→Synthesizer)
• synthesizer_output.json: full protocol with barcoding strategy, timepoints, readouts, pilot plan
• PROTOCOL.md: human-readable experiment protocol

Work Log

2026-04-27 — minimax:72 (iteration 4)

• Final staleness review: confirmed all 4 artifact files on main (commit e89a9cc5e) in analyses/SDA-2026-04-24-allen-shendure-vaccine-lineage-tracing/.
• debate.json: 4 rounds (proposer→jay-shendure→sue-kaech→synthesizer) ✓
• synthesizer_output.json: feasibility_score=0.62, persona_greenlight.jay-shendure=yes, persona_greenlight.sue-kaech=yes, allen_datasets references Seattle Hub for Synthetic Biology and Shendure Lab ✓
• PROTOCOL.md: TracerSeq-Immune v1, VSV-G delivery, dLNGFR reporter, influenza NP mRNA-LNP challenge, timepoints, 3-month pilot plan — all present ✓
• All 5 acceptance criteria verified complete; task is done. No further substantive work required.
• Pushed only the worktree slot file (.orchestra-slot.json) since no artifact changes were needed.

2026-04-27 — claude-sonnet-4-6 (iteration 2)

• Verified core artifact files (debate.json, metadata.json, synthesizer_output.json) from prior iteration present in commit 5ee71c519.
• Added PROTOCOL.md: human-readable protocol covering TracerSeq-Immune v1 barcoding strategy, delivery vehicle, timepoints, sequencing depth, immune challenge model, readouts, persona greenlights, and 3-month pilot plan.
• All 5 acceptance criteria confirmed: 4-round debate, full protocol spec, jay-shendure=yes + sue-kaech=yes, feasibility=0.62≥0.5, Seattle Hub referenced.

2026-04-27 — claude-sonnet-4-6 (iteration 3)

• Final verification: all 4 artifact files present on main at commit e89a9cc5e.
• debate.json: 4 rounds confirmed (proposer→jay-shendure→sue-kaech→synthesizer).
• synthesizer_output.json: feasibility_score=0.62, persona_greenlight.jay-shendure=yes, persona_greenlight.sue-kaech=yes, allen_datasets references Seattle Hub for Synthetic Biology and Shendure Lab.
• PROTOCOL.md: TracerSeq-Immune v1, VSV-G delivery, dLNGFR reporter, influenza NP mRNA-LNP challenge, timepoints, 3-month pilot plan — all present.
• All 5 acceptance criteria verified complete. Task complete on main.

---

Ed Lein × SEA-AD Experiment (task: 7efe7218-8f33-4039-a6f9-f5ecf216064e)

Acceptance Criteria

Artifact

analyses/SDA-2026-04-27-allen-ed-lein-cell-type-vulnerability-ad/

• metadata.json: feasibility=0.62, IIG=0.65, cost=$185K, greenlight=ed_lein yes
• debate.json: 5 rounds (Proposer→Methodologist→Statistician→Ed-Lein-Critic→Red-Teamer)
• synthesizer_output.json: structured protocol with predicted outcomes, endpoints, donor cohort
• PROTOCOL.md: human-readable experiment protocol

Work Log

2026-04-27 06:30 PT — minimax:70 (iteration 3)

• Confirmed Ed Lein × SEA-AD experiment artifact on main at commit cb2ae3b5b.
• Artifact: analyses/SDA-2026-04-27-allen-ed-lein-cell-type-vulnerability-ad/
• All 5 acceptance criteria verified: greenlight=ed_lein yes (feasibility 0.62), SEA-AD cited, pilot plan present (3-month), 5-round debate, IIG/$K=3.5 above class floor.
• Added Ed Lein experiment section to this spec; updated acceptance criteria checkboxes.
• Pushing only spec update — all substantive artifacts already on main.

---

Hongkui Zeng × Mouse Connectivity Atlas Experiment (task: cab084a9-74e5-4a2f-acef-b8fad85a2bf2)

Goal

Propose a showcase experiment Hongkui Zeng's team (Brain Science, Allen Institute) would greenlight, anchored on the Mouse Brain Connectivity Atlas and the Brain Knowledge Platform. The target question: do connectivity motifs predict which cell types in a neurodegenerative mouse model will show first transcriptomic stress signatures? Can we pre-identify vulnerable circuits from anatomy alone?

Acceptance Criteria

Approach

Read the Hongkui Zeng persona (personas/hongkui-zeng/SKILL.md) and Karel Svoboda persona for reviewer lenses.

Research the Allen Mouse Brain Connectivity Atlas projection matrix, the Mouse Whole Brain Cell Type Atlas, the CCFv3, and relevant neurodegenerative mouse models.

Design a connectivity-prior-based experiment that predicts cell-type-specific vulnerability in the 5xFAD mouse model.

Write a four-round debate (proposer → Zeng → Svoboda → synthesizer) and a structured synthesis artifact.

Commit the experiment package and update this spec.

Dependencies

• d40b75de-d60d-485b-9abf-1fb83afeea00 — Hongkui Zeng persona exists on main (commit c8f47796d).
• 26240872-6ffc-40c2-9522-69a812d2f0e1 — Shendure showcase experiment provides the artifact template structure.

Dependents

• Future Allen-experiment tasks can reuse the Zeng experiment artifact as a template for persona-gated showcase experiment proposals focused on connectivity and disease.

Work Log

2026-04-27 — minimax:77 (iteration 8, final verification)

• Final staleness review: confirmed all 4 artifact files present in analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/ in the worktree.
• Verified debate.json (4 rounds: Proposer→Hongkui-Zeng→Karel-Svoboda→Synthesizer) ✓
• Verified synthesizer_output.json: feasibility=0.58, IIG=0.70, cost=$130K, persona_greenlight.hongkui-zeng=yes, persona_greenlight.karel-svoboda=yes, allen_datasets cites Allen Mouse Brain Connectivity Atlas ✓
• Verified PROTOCOL.md: Predicted Outcomes section present (4 scenarios with power calc), 3-month pilot plan, mouse model (5xFAD), transgenic lines (Tlx3-Cre, Pvalb-IRES-Cre), connectivity priors (convergent input, input diversity, reciprocity), primary endpoint (Spearman rho > 0.5, mixed-effects model) ✓
• All 5 acceptance criteria verified complete: 4-round debate ✓, synthesis specifies model/lines/priors/plan ✓, both greenlights yes ✓, feasibility=0.58≥0.5 ✓, allen_datasets cites Connectivity Atlas ✓.
• No further substantive work required; pushing only spec work log update.

2026-04-27 — claude-sonnet-4-6 (iteration 6, predicted_outcomes in JSON)

• Added predicted_outcomes array (4 structured scenarios: Positive, Cell-type-specific, Temporal gradient, Null/falsifiable) directly to synthesizer_output.json — previously only present in PROTOCOL.md.
• synthesizer_output.json now has all required fields: predicted_outcomes, iig_estimate, cost_estimate, pilot_plan, protocol (mouse_model, transgenic_lines, connectivity_priors, timepoints, sequencing, readouts), statistical_model, persona_greenlight (hongkui-zeng=yes, karel-svoboda=yes), allen_datasets (Connectivity Atlas).
• All 5 acceptance criteria confirmed: 4-round debate ✓, synthesis specifies model/lines/priors/plan ✓, both greenlights yes ✓, feasibility=0.58≥0.5 ✓, allen_datasets cites Connectivity Atlas ✓.

2026-04-27 — claude-sonnet-4-6 (iteration 5, predicted outcomes)

• Task description requires "protocol with predicted outcomes" — added explicit Predicted Outcomes section to PROTOCOL.md (4 scenarios: positive, cell-type-specific, temporal gradient, null/falsifiable) with power calculation.
• PROTOCOL.md now fully satisfies all task requirements: predicted outcomes, IIG estimate ($0.70), cost estimate ($130K), 3-month pilot plan, mouse model (5xFAD), transgenic lines (Tlx3-Cre + Pvalb-IRES-Cre), connectivity priors (convergent input, input diversity, reciprocity), primary endpoint (Spearman rho > 0.5, mixed-effects model).
• All 5 acceptance criteria confirmed: 4-round debate, full protocol spec, hongkui-zeng=yes + karel-svoboda=yes, feasibility=0.58≥0.5, allen_datasets includes Connectivity Atlas.

2026-04-27 05:50 PT — claude-sonnet-4-6 (iteration 4, completion verification)

• Verified all experiment artifacts are on main at commit 0e5c73f9a:

- analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/{metadata.json, debate.json, synthesizer_output.json, PROTOCOL.md}

• Confirmed all 5 acceptance criteria:

1. Four-round debate exists (proposer + hongkui-zeng + karel-svoboda + synthesizer) ✓
2. Synthesis specifies: 5xFAD mouse model (JAX #34840), Tlx3-Cre + Pvalb-IRES-Cre × INTACT2, 3 connectivity priors (convergent input, input diversity, reciprocity), timepoints 2/3/4 months, 50K reads/nucleus, mixed-effects model ✓
3. 3-month pilot plan present; hongkui-zeng=yes, karel-svoboda=yes (conditional on c-Fos, included) ✓
4. feasibility_score=0.58≥0.5; Allen Mouse Brain Connectivity Atlas cited (Oh et al., Nature 2014, PMID: 25470075) ✓
5. allen_datasets field includes Allen Mouse Brain Connectivity Atlas ✓

• Checked all acceptance criteria boxes to reflect completion status.

2026-04-25 16:58 PT — glm-5:63

• Started staleness review from worktree at 8863c62c8 (HEAD of main). Rebased onto latest origin/main.
• Confirmed Hongkui Zeng persona exists on main (commit c8f47796d) and recovered content via git show origin/main:personas/hongkui-zeng/SKILL.md. Persona not present in worktree checkout but accessible from main.
• Confirmed Karel Svoboda persona exists on main.
• Reviewed existing Shendure showcase experiment (SDA-2026-04-24-allen-shendure-vaccine-lineage-tracing) as template for artifact structure: metadata.json, debate.json, synthesizer_output.json.
• Gathered scientific background: Allen Mouse Brain Connectivity Atlas (Oh et al., Nature 2014, PMID: 25470075), Mouse Whole Brain Cell Type Atlas (Yao et al., Nature 2023, PMID: 37656950), CCFv3 (Wang et al., Cell 2020, PMID: 32386544), 5xFAD model, Allen Cre driver lines (Tlx3-Cre, Pvalb-IRES-Cre), INTACT2 nuclear tagging.
• Designed a connectivity-prior-based experiment: compute convergent-input, input-diversity, and reciprocity metrics from the Allen projection matrix; select 6 predicted-vulnerable and predicted-resilient circuits; test with cell-type-specific snRNA-seq in 5xFAD mice.
• Created 4-round debate: proposer → hongkui-zeng (greenlight yes with refinements) → karel-svoboda (greenlight yes with activity control) → synthesizer (final protocol).
• Created structured synthesis with: 5xFAD mouse model, Tlx3-Cre + Pvalb-IRES-Cre lines, INTACT2 nuclear tagging, 72 snRNA-seq libraries, c-Fos and 6E10 IHC controls, 3-month pilot plan, $130K cost estimate, IIG 0.7, feasibility 0.58.
• Both persona greenlights achieved: hongkui-zeng = yes, karel-svoboda = yes (conditional on c-Fos activity control, which is included).
• Artifact: analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/

2026-04-26 00:50 PT — glm-5:63 (iteration 3, gate-fix retry)

• Previous iteration rejected by review gate: branch was contaminated with unrelated file deletions (api.py, debate_fanout.py, etc.) from a bad rebase.
• Reset branch to origin/main, re-added only the PROTOCOL.md artifact.
• Verified all 5 acceptance criteria pass: 4-round debate exists, synthesis specifies model/lines/priors/plan, both persona greenlights = yes, feasibility 0.58, allen_datasets cites Connectivity Atlas.
• Added PROTOCOL.md: human-readable experiment protocol summary with connectivity priors, experimental design, controls, statistical model, 3-month pilot plan, and cost estimate.

2026-04-26 01:04 PT — glm-5:63 (iteration 4, gate-fix retry 2)

• Worktree was clean at origin/main; no stale local changes to api.py or other unrelated files.
• Verified all experiment artifacts on main: analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/{debate.json, metadata.json, synthesizer_output.json, PROTOCOL.md}.
• Verified all 5 acceptance criteria: (1) 4-round debate ✓, (2) synthesis has 5xFAD model + Tlx3-Cre/Pvalb-IRES-Cre lines + 3 connectivity priors + 2/3/4mo timepoints + 50K reads/nucleus + mixed-effects model ✓, (3) 3-month pilot plan + hongkui-zeng=yes + karel-svoboda=yes ✓, (4) feasibility=0.58≥0.5 + cites Connectivity Atlas ✓, (5) allen_datasets includes Connectivity Atlas ✓.
• This iteration pushes only the spec work log update. All substantive experiment artifacts are already on main (commits 2ec2092e4, 2bdfbd097, 8c52aaec5).

2026-04-27 — claude-sonnet-4-6 (iteration 4, final verification)

• Re-verified: all 4 artifact files present on main at commit 0e5c73f9a (analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/).
• Confirmed synthesizer_output.json: feasibility=0.58, IIG=0.70, cost=$130K, hongkui-zeng=yes, karel-svoboda=yes, allen_datasets includes Allen Mouse Brain Connectivity Atlas.
• All 5 acceptance criteria confirmed complete; no further substantive work required. Task is done.

2026-04-25 23:58 PT — glm-5:63 (iteration 2)

• Picked up iteration 2. Verified iteration 1 commit 3a01433cf present on branch with complete artifact.
• Validated all 5 acceptance criteria against the artifact: all met.
• Rebased on latest origin/main (already up to date). Pushing for validator review.

2026-04-27 06:20 PT — minimax:70 (iteration 3)

• Confirmed Ed Lein × SEA-AD experiment artifact is already on main (commit cb2ae3b5b).
• Artifact: analyses/SDA-2026-04-27-allen-ed-lein-cell-type-vulnerability-ad/
• All acceptance criteria verified: greenlight=ed_lein yes (0.62 feasibility), SEA-AD cited, pilot plan present, 4-round debate (Proposer→Methodologist→Statistician→Ed-Lein→Falsifier), IIG=0.65, cost=$185K, IIG/$K=3.5 above class floor.
• JSON validation found two syntax errors in synthesizer_output.json (inline Python expressions and inline comments in JSON) — fixed both.
• Fixed: (1) sample_size values had inline comments removed to proper separate notes field; (2) iig_per_dollar had arithmetic expression removed, stored as string.

2026-04-27 05:45 PT — claude-sonnet-4-6 (iteration 3, clean recreation)

• Previous iterations' branch was rejected by merge gate due to unrelated file deletions (api.py, debate_fanout.py, etc.) from bad rebase contamination in an earlier worktree.
• This iteration starts fresh from origin/main (commit 43d50fc6c). Previous artifacts existed only in rejected branch history.
• Recreated all 4 experiment artifact files by recovering content from prior iteration git history:

- analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/metadata.json
- analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/debate.json (4 rounds: proposer → hongkui-zeng → karel-svoboda → synthesizer)
- analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/synthesizer_output.json
- analyses/SDA-2026-04-25-allen-zeng-connectivity-vulnerability-circuits/PROTOCOL.md

• All 5 acceptance criteria verified:

1. Four-round debate artifact exists (proposer + 3 personas) ✓
2. Synthesis specifies: 5xFAD mouse model, Tlx3-Cre + Pvalb-IRES-Cre, 3 connectivity priors (convergent input, input diversity, reciprocity), timepoints 2/3/4 months, 50K reads/nucleus, mixed-effects statistical model ✓
3. 3-month pilot plan present; hongkui-zeng=yes, karel-svoboda=yes (conditional on c-Fos, which is included) ✓
4. feasibility_score=0.58≥0.5; Allen Mouse Brain Connectivity Atlas cited (Oh et al., Nature 2014, PMID: 25470075) ✓
5. allen_datasets field includes Allen Mouse Brain Connectivity Atlas ✓

• Staged only the 4 new artifact files + this spec update. api.py and all other critical files untouched.

2026-04-24 19:55 PT — Codex (Shendure experiment, task 26240872)

• Started staleness review from the current worktree at 8ae0c20e4; confirmed no existing in-tree artifact covers this Seattle Hub adaptive-immunity proposal.
• Reviewed local analysis artifact structure and recovered Jay Shendure persona guidance from recent commit 31c3da6cf because personas/jay-shendure/SKILL.md is not present in this checkout.
• Gathered supporting references from Allen Institute pages for the Seattle Hub / Shendure lab and from PubMed for molecular recording, vaccine memory, and recall repertoire dynamics.
• Chose a CD8-memory-centered pilot because it best fits Sue Kaech's domain lens and gives a clean challenge-response phenotype.
• Created a new analysis package with debate + synthesis (SDA-2026-04-24-allen-shendure-vaccine-lineage-tracing).