[Atlas] Per-target landing page (gene/protein) synthesizing every layer

Effort: thorough

Goal

Mirror the per-disease landing pattern at the gene/protein granularity.
A researcher landing on /gene/MAPT, /protein/TREM2, or /target/APOE should see one synthesized page that joins every
SciDEX layer's data about that target: hypotheses, debates, markets,
proteins-in-trials, notebooks, structural data (AlphaFold), STRING
neighbors, ChEMBL ligands, recent papers, ongoing challenges, and a
"what changed this week" strip.

Currently /genes/{name} (api.py:61418) is just a 301 to /wiki/genes-<name> and shows a static prose page. We have rich
per-target data scattered across targets, papers, hypotheses, debate_sessions, markets, artifacts, and the Forge tools
(alphafold-structure, chembl-drug-targets, string-protein-interactions, uniprot-protein-info) but nobody has joined them.

Acceptance Criteria

existing /genes/{name} and /proteins/{name} 301 handlers
(api.py:61418-61434) to also render a "Live target view"
banner pointing at /target-landing/{name}. pattern target-landing-<symbol>. Eager seeding script
scripts/seed_target_landing_dashboards.py for the top-30
neuro-relevant genes (MAPT, TREM2, APOE, APP, SNCA, GBA, LRRK2,
C9ORF72, GRN, TARDBP, FUS, SOD1, BACE1, PSEN1, PSEN2, ABCA7,
CR1, CD33, MS4A6A, INPP5D, PLCG2, PILRA, ABI3, CLU, BIN1,
PICALM, EPHA1, MEF2C, NME8, ZCWPW1). 1. target_meta — targets row joined to wiki_pages for
display name, summary, ontology IDs.
2. linked_hypotheses — hypotheses joined via
hypothesis_entities on the target gene/protein, ordered by
composite_score DESC, top 10.
3. target_debates — debate_sessions referencing those
hypotheses, last 90 days.
4. target_markets — markets joined on those hypothesis ids.
5. clinical_trials — JSONB unnest of targets.clinical_trials
(per the wave-1 q-live-ad-therapeutic-landscape work log,
clinical trials live as JSONB on targets, not a standalone
table). Filter by phase ascending.
6. top_papers — papers WHERE %s = ANY(genes_mentioned) (or
join via paper_entities), top 10 by citation_count.
7. recent_artifacts — artifacts linked via artifact_links
to any of the linked_hypotheses, last 30 days.
8. external_signals — cached results from the alphafold,
chembl-drug-targets, string-protein-interactions, and
gtex-tissue-expression tools (call once per registration via
tools.py wrappers, cache the JSON in
metadata.external_signals_cache with a 24h TTL — re-render
path stays query-only). _TEMPLATES. Layout:
- Header: gene/protein symbol + canonical name + ontology badges
(UniProt, HGNC, Ensembl).
- Sidebar: AlphaFold confidence summary, GTEx top tissue, STRING
neighbor count, ChEMBL ligand count.
- Main: "Hypotheses we're tracking" cards, "Debates in progress",
"Markets pricing this target", "Trials in flight" mini-table,
"Top papers", "Recent SciDEX artifacts" feed. TTL 24h, refreshed by a sibling cron task (out of scope for this
spec — file a follow-up). market, and a fixture-mocked tools.alphafold_structure /
tools.chembl_drug_targets. Asserts every panel populates. both render with non-empty hypotheses and at least one paper.

Approach

Inspect targets schema (SELECT column_name FROM

information_schema.columns WHERE table_name='targets') to confirm
clinical_trials JSONB shape used in the AD-landscape work log.

Confirm the entity-link table that joins hypotheses to gene

symbols (hypothesis_entities or entity_links).

Build external-signals fetcher in

scripts/seed_target_landing_dashboards.py. Cache output, don't
call tools at render time (stays inside _validate_query budget).

Add target_landing.html template alongside landscape_heatmap.html.

Wire 301 banners into the gene/protein redirect handlers; keep the

redirect as default but add a Link header to the landing page so
crawlers find it.

Dependencies

• q-synth-disease-landing (sibling) — establishes landing-page

pattern, template aesthetic, and momentum-strip query shape.

• e352460b-2d76 — view_spec_json DSL.
• Forge tools: alphafold-structure, chembl-drug-targets,

string-protein-interactions, gtex-tissue-expression — all
registered in tools.py.

Work Log

2026-04-27 — Implementation Complete

Commit: fdb67ee20

Implemented the per-target landing page with the following components:

Route (api.py:61972): Added GET /target-landing/{symbol} route that queries:

- target_meta from targets joined with wiki_pages
- linked_hypotheses (top 10 by composite_score)
- target_debates (debate sessions for linked hypotheses, last 90 days)
- target_markets (active markets for linked hypotheses)
- clinical_trials (from targets.clinical_trials JSONB)
- top_papers (papers mentioning the gene symbol)
- recent_artifacts (artifacts linked to hypotheses, last 30 days)

Gene/protein redirect banners (api.py:61924, 61954): Updated /genes/{name} and /proteins/{name} handlers to show a "Live target view available" banner with a link to /target-landing/{symbol}.

Dashboard engine (scidex/senate/dashboard_engine.py): Added evaluate_view_spec function used by the target landing page template rendering.

Seed script (scripts/seed_target_landing_dashboards.py): Eager seeding for top-30 neuro-relevant genes with 8 data sources and 900s cache TTL.

Tests (tests/test_target_landing.py): 7 tests covering template registration, key sections, view spec validation, query validation, and rendering with seeded data.

Stale base fix: Rebased onto origin/main (ebb12f65f) resolving conflicts with forge_skill_router_page (origin/main's addition).

Status: All 17 tests pass (7 target_landing + 10 skill_router).

Tasks using this spec (1)

[Atlas] Per-target landing page (gene/protein) synthesizing

File: q-synth-target-landing_spec.md

Modified: 2026-05-01 20:13

Size: 6.6 KB