Discover causal drug targets for idiopathic pulmonary fibrosis from public GWAS data

Using a genetics-first pipeline (GWAS → TWAS → colocalization → Mendelian randomization → druggability → safety), identify and rank novel causal drug targets for Idiopathic Pulmonary Fibrosis (IPF) entirely from public data and computational methods. IPF was selected for its strong GWAS architecture (~27 loci), high unmet need (only 3 approved drugs, none with genetic support), and clear tissue context (lung). The core scientific question: do genetically-supported targets overlap with what's in clinical trials — and what is pharma missing? All work lives in /workspace/drug-target-discovery/; bioinformatics only, no lab experiments, public/published data only.

Sessions

• Session 1 (2026-03-12): Selected IPF as target disease; curated 27 GWAS loci across 6 publications with verified GRCh38 coordinates. Full log

• Session 2 (2026-03-12): Built full pipeline in one pass — TWAS (15 genes), colocalization (13 results), MR (9 genes), initial 20-target ranking; found ~80-85% of IPF clinical pipeline lacks genetic support. Full log

• Session 3 (2026-03-12): Verified all 27 GWAS SNP positions against dbSNP; expanded to 21 TWAS entries, 18 coloc entries, 13 MR entries; discovered ATP11A strong coloc (PP.H4=0.996) and DPP9 fibroblast-specific coloc. Full log

• Session 4 (2026-03-12): Queried DepMap 25Q3 API (real Chronos scores for all 21 targets — none common essential; TERT reclassified to strongly selective) and Open Targets GraphQL API (ACVRL1 most tractable: 10 buckets); added 3 new GWAS loci from Chin 2025 (MUC1, NTN4, SLC6A6; now 30 loci total); cross-validated against rentosertib Phase 2a (+98.4 mL FVC), bexotegrast failure, SPL5B MUC5B ASO Phase 1, and DPP9 inhibitor compound correction (42/47 not 6e, 175x selectivity). Full log

• Session 5 (2026-03-26): Comprehensive translation roadmap completed — integrated 2026 clinical trial updates (TETON-2 published +95.6 mL FVC, rentosertib Nature Med +98.4 mL, Compound 6e oral bioavailable, ALOFT-IPF Phase 3 completion Oct 2026, CYT107 >650 patients dosed safely); created clinicaltranslationroadmap.md (Tier 1 immediate actions: IL-7 Phase 2a, sirolimus Phase 2b, DPP9 paradox resolution); created ipfpathwayintegrationmap.md (6 biological pathways, all 21 targets integrated, key convergence points identified); identified three critical opportunity gaps (IL-7, sirolimus, DPP9 all untested in IPF despite strong evidence and available drugs); resolved DPP9 therapeutic paradox via dual mechanism (inflammasome + FAP-EMT). Full log

• Session 6 (2026-03-26): DPP9 paradox deepened — TC-E5007 is anti-fibrotic in kidney via TGF-β1/Smad (PMID:33932609), pinocembrin identified as first DPP9 activator, Cpd42 has only 2% bioavail while ICeD-2 has 61%; MUC1 upgraded to Tier 1 rank 8 (KO mice protected from fibrosis, GO-201 anti-fibrotic in vivo; Thorax PMID:31801904); SPDL1 validates spindle pathway; zelasudil ROCK inhibitor Phase 2a provides first genetic-clinical convergence via AKAP13; pipeline expanded to 25 targets. Full log

• Session 7 (2026-03-26): Created standalone DPP9 therapeutic analysis with Harapas 2022 IL-1R haploinsufficiency rescue as key evidence (removing 1 copy of Il-1r rescues DPP9-null mice → proves partial anti-IL-1β blockade suffices for IPF); defined NLRP1 vs CARD8 mechanistic distinction; added DPP9-KEAP1-NRF2 redox axis; updated AKAP13 with zelasudil Phase 2a hold lifted; upgraded pirfenidone genetic evidence to "Moderate" via MUC1-CT mechanism link. Full log

• Session 8 (2026-03-26): Synthesized DPP9 "double-hit" model (NLRP1 inflammasome + KEAP1-NRF2 redox) with 5 therapeutic strategies ranked; added MCL1 protective locus from eBioMedicine 2025 WGS + NTN4 sQTL lung tissue evidence + talabostat antifibrotic preclinical data; expanded to 33 GWAS loci and 25 colocalization entries; pipeline now at 25 targets across 6 pathway clusters. Full log

• Session 9 (2026-03-26): Created comprehensive pathwaynetworkanalysis.md mapping 25 targets across 6 pathways with 3 convergence nodes and 8 ranked drug repurposing opportunities; independently verified IL-1β translational gap (zero IPF trials for anakinra/canakinumab); characterized NTN4 biology comprehensively (BM disruption via laminin γ1, anti-senescence, 26.6x lung-enriched); validated DPP9-KEAP1-NRF2 axis with additional sources (Cancer Res 2023 ferroptosis, DiPAK sensor 2025); updated clinical gap analysis with pathway convergence table. Full log

• Session 10 (2026-03-26): Captured deupirfenidone Phase 2b quantitative data (FVC -21.5 mL near-stabilization, HR=0.439 p=0.0023, 98.5% posterior probability; Phase 3 SURPASS-IPF starting H1 2026 via 505(b)(2)); confirmed TETON-1 imminent (TETON-2 95.6 mL in NEJM March 2026 — largest IPF Phase 3 effect); verified Compound 6e citation (J Med Chem 68:23163-23184); synthesized 7-cluster convergence diagram into clinicalgapanalysis.md; revised TOP 5 actionable targets with MUC1-CT #1. Full log

• Session 11 (2026-03-26): Integrated hepatocyte-specific DPP9 KO tolerance data (tissue-specific loss survivable, autophagy dominant over inflammasome; ScienceDirect 2025); created formal DPP9 4-strategy therapeutic framework ranking anti-IL-1β as primary recommendation; built deep MUC1 functional profiles (T1224/Y1229 phosphorylation, GO-203 nanoparticles, pirfenidone MUC1-CT mechanism, galectin-3 bridge); characterized NTN4 as biologically distinct from NTN1 (laminin β1 not γ1 homology; no shared receptors); added SLC6A6 mitochondrial taurine import biology (NatMetab 2026); identified DPP9 as highest-connectivity network hub (3/6 clusters); updated clinical trials with TETON-2 specific numbers (ΔFVC +95.6 mL, HR=0.71), admilparant Phase 2 (+1.4% FVC), TPIP Phase 3 launch, zelasudil hold lifted. Full log

• Session 12 (2026-03-26): Deep structural analysis of DPP9-NLRP1 ternary complex (cryo-EM PDB 6X6A/7CRW; ZU5 "bomb defuser" mechanism); corrected DPP9 fibroblast coloc to published 0.968 (Dalapati 2025 PMID:39876559); filled Chin 2025 GWAS loci with exact ORs (MUC1 1.162, NTN4 1.153); established NTN4 has completely different receptor biology from NTN1 (neogenin not DCC/UNC5; expressed in aberrant basaloid cells); documented 3/5 recent IPF trial failures lacked genetic support (bexotegrast, GB0139, zinpentraxin — 60% failure rate without genetics). Full log

• Session 13 (2026-03-27): Deep-dives on DPP9 (CARD8 threshold mechanism, pinocembrin as first activator, tissue-specific paradox: kidney inhibition antifibrotic vs lung loss pathogenic), MUC1 (Natri 2024 sc-eQTL confirms AT2 cell coloc across 3 GWAS, ADAM17/EZH2 axis, dual role caveat), NTN4 (26.6x lung-enriched, aberrant basaloid cells, anti-senescence/anti-inflammatory, no fibrosis study exists); expanded pathway network to 10 clusters with real PPIs (BTRC-RIPK4 direct ubiquitination, AKAP13-FAM13A opposing RhoA control); quantified cross-ancestry allele frequencies (MUC5B 50-fold EUR/AFR differential, non-telomere IPF-PRS portable, TERT/DSP most globally relevant). Full log