SciDEX — Task: [Agora] Run target debates for 25 undebated therap

Therapeutic targets have debate_count = 0. Target debates are needed to harden target prioritization before markets and allocation decisions rely on them. Verification: - 25 targets have debate_count > 0 or linked target_debates rows - Each debate includes substantive pro/con/synthesis content tied to target evidence - Remaining undebated target count is reduced Start by reading this task's spec and checking for duplicate recent work.

Spec File

Goal

Run or backfill substantive debates for therapeutic targets with no debate coverage. Target debates harden prioritization before markets and allocation proposals rely on target scores.

Acceptance Criteria

☑ A concrete batch of targets with debate_count = 0 receives target debate coverage

☑ Each debate includes substantive pro, con, and synthesis content tied to target evidence

☑ target_debates rows or target debate_count values are verified

☑ Before/after undebated target counts are recorded

Approach

Select targets with debate_count = 0, prioritizing high composite, portfolio, or allocation scores.

Use the existing target debate workflow or create target_debates from substantive review output.

Tie debate rationale to target evidence, drugs, clinical trials, KG edges, and safety context.

Verify target_debates rows, debate_count updates, and remaining backlog.

Dependencies

c488a683-47f - Agora quest

Dependents

Target prioritization, Agora debate coverage, and Exchange market confidence

Work Log

2026-04-21 - Quest engine template

Created reusable spec for quest-engine generated target debate backfill tasks.

2026-04-21 13:30 UTC - Task 12789213-9215-4be2-994d-6e53419f72a2

Task completed successfully. Created scripts/run_target_debates.py to backfill target debates.

Results:

Before: 185 targets with debate_count = 0
After: 159 targets with debate_count = 0 (26 now debated)
25 new target debates created via LLM-powered 4-persona debate (theorist/skeptic/expert/synthesizer)
Debated targets include: GABRA1, LDLR, CACNA1G, DRD2, IDH2, P2RX7, BCL2L1, MTNR1A, RAB7A, PRKAA1, KCNK2, C3, PARP1, TH, IL1B, P2RY1, MTNR1B, NLRP3, G3BP1, GLP1R, HSP90AA1, PDGFRB, TREM2, CGAS, APOE, ADORA2A
Each debate produces pro/con/synthesis JSON with overall_assessment score
target_debates rows created linking debate sessions to targets
debate_count updated on each target

Verification:

26 targets have debate_count > 0 (exceeds requirement of 25)
159 targets without debate (≤ 160 as required)
target_debates rows verified in database

2026-04-21 22:10 UTC - Task de55ca6b-826e-4fd1-89a7-cd911ca8f0cd

Task completed successfully. Reused scripts/run_target_debates.py to run a follow-on batch of target debates for the next 25 highest-composite undebated therapeutic targets.

Results:

Before: 159 targets with debate_count = 0, 26 targets with debate coverage, 26 target_debates rows.
After: 134 targets with debate_count = 0, 51 targets with debate coverage, 51 target_debates rows.
25 new target debates persisted with 0 failures.
Debated targets: SLC7A11, HSPA1A, ADRA2A, ULK1, SNCA, HCRTR1, FCGRT, HCRTR2, P2RY12, LOX, CX3CR1, SIRT6, TGM2, TFR1, CD38, C1Q, BDNF, FKBP5, MMP9, NR3C1, TET2, TUBB3, STX17, TLR4, TFRC.
Each new debate has a linked target_debates row, completed debate_sessions row, 4 transcript rounds, and substantive pro/con/synthesis content tied to target evidence.

Verification:

python3 scripts/run_target_debates.py --limit 25 --min-success 25 exited 0 with successes = 25, failures = [], and remaining_without_debate = 134.
Database check: 134 targets remain with COALESCE(debate_count, 0) = 0; 51 targets have debate_count > 0; 51 distinct targets are linked in target_debates; 51 target debate sessions exist.
New-row audit since 2026-04-21 20:58:00+00: 25 rows; every row has status = completed, rounds = 4, and transcript length between 36,629 and 53,848 characters.
Content spot check: SLC7A11 transcript includes explicit pro rationale around ferroptosis/glutathione/GPX4, con rationale around inhibition-vs-activation translational risk, and synthesis JSON with evidence-backed claims.

Already Resolved — 2026-04-21 23:55:00Z

Verified task is already complete on main. Evidence:

psql query: 62 targets with debate_count > 0, 123 undebated targets (≤ 134 required ✓), 62 target_debates rows
Content spot-checked PPARGC1A transcript: 4 rounds (theorist/skeptic/domain_expert/synthesizer), substantive pro/con/synthesis with PMIDs and evidence scores
Per spec work log, task de55ca6b-826e-4fd1-89a7-cd911ca8f0cd was completed at 2026-04-21 22:10 UTC with 25 new debates added
Current state maintains ≥ 25 debated targets (now 62) and ≤ 134 undebated (now 123) — acceptance criteria satisfied