Built with Claude · Life Sciences · Researcher track

Selective Th2 Suppressor Re-analysis

A genome-scale CRISPRi Perturb-seq re-analysis in primary human CD4+ T cells — hunting knockdowns that suppress the Th2 (allergy) program without flipping cells toward Th1, stress-testing that claim to destruction, then asking whether a different, degrader-relevant phenotype is the more useful question.

Honest bottom line. On the core claim — selective Th2 suppression — the answer is a calibrated negative, reached independently by three tracks: competitive scoring collapses the arm-correlation confound (+0.54→+0.18), a permutation null shows no enrichment over random gene sets (FDR ≈ 1), and GATA3 reads as a Th1-skewer under the matched Th1-vs-Th0 arm. A fourth track asks a different question — not "selective," but "phenocopies the STAT6/GATA3 collapse," the bar an oral STAT6 degrader actually clears. That reframe calibrates as a classifier (anchors self-recover, permutation p < 0.001) and, fused with Open Targets tractability + AD genetics + AD patient skin, nominates druggable off-axis nodes (ITK, INPP5D/SHIP1) — a nomination map, not a validated target list, and it does not rescue the selectivity negative. See RECONCILIATION.md.
Interactive report

📊 Dashboard

The full self-contained story: the 2D arm plane, scored shortlist, mechanism dossiers, independent review, and the hardened re-analysis — all readable in-page.

Open the dashboard →
Raw artifacts

🗂️ File index

Click through every underlying file — figures, tables, parquet data, dossiers, and the analysis notebooks — with relative links.

Browse the files →
Independent replication · primer

🧭 Independent re-analysis

A from-scratch replication — streaming pipeline, its own QC catch, independent verification — that reaches the same negative result, plus a biology primer and the full reasoning as an interactive page.

Open the primer →
Next-degrader strand · slide deck

💊 Druggable type-2 network

A degrader-relevant reframe: score every perturbation for STAT6/GATA3 collapse phenocopy (a positive anchor that calibrates where the selectivity residual gave FDR ≈ 1), then weight by chemical tractability, AD genetics, and AD patient skin. Nominates off-axis nodes — ITK & INPP5D/SHIP1 — conditioned on the same negative.

Open the deck →
Next-degrader strand · methods

📁 Druggable-network project

The reusable collapse scorer, the 63-node ranked map, the patient-anchoring layer, and a report — three-notebook reproducible pipeline.

Read the project on GitHub →

What's inside

Next-degrader strand — the druggable type-2 network

Data: Zhu & Dann et al. 2025, genome-scale CRISPRi Perturb-seq in primary human CD4+ T cells (Marson lab); signature: Ota et al. 2021. See the repository README for full provenance and citations.
View source on GitHub →