Bench to Biobank
Autoimmune drug discovery often fails when a gene that looks important in a cell experiment doesn't hold up in real people.
Core trick
Each regulator passes sequential evidence filters from CRISPRi knockdown through GWAS, PheWAS, and individual-level All of Us genomes, with nearest-gene stress tests.
Description
Autoimmune drug discovery often fails when a gene that looks important in a cell
experiment doesn't hold up in real people. We built a reproducible pipeline that
takes drug-target hypotheses from the Marson-lab genome-scale CD4+ T-cell
Perturb-seq screen all the way to individual-level human genomes in All of Us.
Every regulator passes through five independent evidence layers: bench causality
(CRISPRi knock-down that calms T-cell activation) → druggability → human GWAS →
biobank PheWAS (FinnGen + GWAS Catalog, cross-ancestry check in BioBank Japan) →
individual-level genomes in All of Us. From ~4,000 regulators, 2,771 pass the
bench filter and 86 are druggable and disease-linked.
What we found: the method works (PTPN22 and TYK2 controls replicate); IL23R
validates for inflammatory bowel disease in the right disease (OR 0.73,
P 4.4x10^-5); the three novel lupus receptor hypotheses return honest, well-
controlled results; and the pipeline catches its own false positives with a
nearest-gene stress test (e.g. our flagship "hidden hit" maps to a famous
neighbor, not our gene). It also surfaced ATP2B1 as an SLE-protective lead, then
honestly down-weighted it when its genetics turned out cardiometabolic.
Why it matters: the discrete result is not "three lupus targets" — it's a
reproducible bench-to-biobank pipeline that confirms a real target, gives honest
nulls, and catches its own false positives. It's packaged as an open, config-
driven tool (bench2biobank) that runs on any perturbation screen. Built entirely
in Claude Science.
README excerpt
# Bench → Biobank: human-genetics-validated CD4⁺ T-cell drug targets for autoimmunity **Built with Claude: Life Sciences** (Jul 7–13, 2026) · *Researcher Track — Build From the Bench* 🔗 **Live interactive demo:** https://kannyjyk.github.io/bench2biobank/ — *(opens the demo with live 3D target structures, no download needed)*  *Every CD4⁺ regulator, tested in* All of Us *individual-level genomes: method controls fire (PTPN22, TYK2), IL23R→IBD confirms (OR 0.73, P = 4.4×10⁻⁵), and the novel bench hypotheses return honest nulls.* > **In one sentence.** Starting from the Marson-lab genome-scale CD4⁺ T-cell Perturb-seq atlas, we > pass every regulator through **five independent evidence layers — bench causality → druggability → > human GWAS → biobank PheWAS → *All of Us* individual-level genomes** — and report what survives each. > The individual-level test is the moat: it **validates the pipeline** (canonical controls fire), > **confirms IL23R→IBD** (protective, P = 4.4×10⁻⁵), returns an **honest null** for the lupus receptor > hypotheses, and — testing "hidden hits" with no common-variant footprint — **surfaces a novel one > (ATP2B1), then down-weights it honestly** when its whole genetic phenome turns out cardiometabolic. ## Why this is a real finding, not just a screen The discrete, defensible result is **not** "three lupus targets." It is a **reproducible bench→biobank pipeline that, tested at indiv