> 매일 아침 research-watcher가 27개 query family × PubMed/GEO를 스캔해서, Brown Biotech 파이프라인 관점에서 decision-ready 인사이트로 정리합니다. 매일 06:00 KST 자동 생성. **Category focus:** Longevity & Senolytics **Published:** 2026-07-20 06:16 KST **Entry ID:** 48 **Tags:** #DeepSAS #deep graph representation learning #heterogeneous graph #contrastive learning #senotype #cellular senescence #senescence-associated genes #longevity #IPF #idiopathic pulmonary fibrosis #CTHRC1+ fibroblast #NFE2L2 --- ## 🔬 Today's Top Findings ### 1. DeepSAS maps IPF senotypes with graph learning and spatial validation (GSE331432, score 6, n=2, Homo sapiens, Xenium spatial transcriptomics, pdat 2026/05/29) ### 2. Barcoded scATAC-seq links CAR-T engineering choices to efficacy-associated chromatin states (GSE331391, score 6, n=13, Homo sapiens + Mus musculus, pdat 2026/07/15) ### 3. TGF-beta-induced SMAD2/3 and YAP/TAZ chromatin occupancy in human fibroblasts (GSE338387, score 6, n=32, Homo sapiens, ChIP-seq, pdat 2026/07/16) ## 📋 Synthesis The 2026-07-20 06:00 KST research-watcher scan completed successfully with 105 hits across 27 queries. Three actionable signals survived the peptide / AI-infrastructure / longevity / cost screen, and all three are NOVEL versus yesterday's id:47 digest and absent from the retained id:43-id:47 window. (1) Longevity + AI drug discovery: GSE331432 introduces DeepSAS, a heterogeneous-graph, attention-based contrastive-learning framework for resolving rare, cell-type-specific senescent cells and senescence-associated genes. The GEO summary reports 1,678 senescent cells among 24,125 cells and 263 senescence-associated genes across 26 cell types in IPF data, with Xenium and human precision-cut lung-slice validation nominating NFE2L2 in CTHRC1+ fibroblasts; the deposited accession itself contains only two hPCLS Xenium samples, so this is a high-value but small-n validation set. (2) Refractory cancer + screening infrastructure: GSE331391 uses genetically encoded barcodes and pooled scATAC-seq to connect targeted CAR architecture and culture-state manipulations to single-cell chromatin outcomes in vitro and after transfer into leukemia-bearing mice. The linked Research Square preprint is PMID 42427866, DOI 10.21203/rs.3.rs-9859689/v1. (3) Fibrosis target validation: GSE338387 provides 32 human vocal-fold-fibroblast ChIP-seq samples comparing TGF-beta1-treated and untreated cells to map SMAD2/3 and YAP/TAZ occupancy. It is a tractable public chromatin dataset for testing convergence between canonical TGF-beta signaling and mechanotransduction, but the GEO record has no linked PMID or DOI and does not by itself establish therapeutic efficacy. Why this matters for Brown Biotech: together these datasets define a reusable pipeline from AI-based senotype discovery, through perturbation-linked single-cell regulatory readouts, to low-cost transcription-factor occupancy validation for IPF/fibrosis and refractory-cancer programs. --- ## 🎯 Highlights ### 1. Longevity + AI drug discovery — GSE331432 (score 6, n=2 deposited Xenium samples; pdat 2026/05/29) reports DeepSAS, which integrates intracellular transcriptional state and intercellular communication in a heterogeneous graph with attention-based contrastive learning. The study-level GEO summary reports 1,678 senescent cells out of 24,125 IPF cells, 263 senescence-associated genes across 26 cell types, and spatial/ex vivo support for NFE2L2 enrichment in CTHRC1+ fibroblasts. Why it matters for BB: this is a concrete senotype-to-target-nomination workflow for the longevity and IPF lanes, but the n=2 deposited validation set requires cross-cohort replication before prioritizing NFE2L2. Source: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE331432. No linked PMID or DOI is listed in GEO. ### 2. Refractory cancer + biotech infrastructure — GSE331391 (score 6, n=13; pdat 2026/07/15) combines genetically encoded CAR barcodes with pooled human and murine scATAC-seq. The linked preprint reports stable and transient transcription-factor activities programmed by cytokine concentration during expansion and altered in-vivo effector differentiation after changes to the CAR antigen-binding domain. Why it matters for BB: it is a reusable screening pattern for ranking CAR or binder designs by downstream regulatory state rather than abundance alone, directly supporting refractory-cancer and perturbation-analysis services. Sources: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE331391; PMID 42427866; DOI https://doi.org/10.21203/rs.3.rs-9859689/v1. ### 3. Fibrosis + target validation — GSE338387 (score 6, n=32; pdat 2026/07/16) maps SMAD2/3 and YAP/TAZ chromatin occupancy in TGF-beta1-treated versus untreated HOVX human vocal-fold fibroblasts using ChIP-seq. Why it matters for BB: the public BW and narrowPeak outputs make this a comparatively low-cost way to test whether canonical TGF-beta and mechanotransduction programs converge at shared regulatory elements, then compare those elements against IPF fibroblast states. Source: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE338387. GEO lists no PMID or DOI, so conclusions should remain at the occupancy/target-validation level. ### 4. Four-axis decision gate — DeepSAS/IPF: peptide 0/3, AI infrastructure 3/3, longevity 3/3, low-cost/ease 3/3, translational fit 3/3 = 12/15; barcoded CAR-T scATAC: 1/3, 3/3, 0/3, 2/3, 3/3 = 9/15; TGF-beta–SMAD/YAP ChIP-seq: 0/3, 1/3, 2/3, 3/3, 2/3 = 8/15. All three are first-time primary features today. Next action: reproduce DeepSAS senotype calls across external IPF cohorts, benchmark barcode-to-state attribution for CAR variants, and intersect SMAD/YAP peaks with CTHRC1+ fibroblast regulons before any chemistry or therapeutic claim. --- ## 💼 Next Steps - **[View DeepSAS / IPF senotype dataset](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE331432)** - **[Read barcoded CAR-T scATAC preprint](https://doi.org/10.21203/rs.3.rs-9859689/v1)** - **[View TGF-beta–SMAD/YAP ChIP-seq dataset](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE338387)** - **[Request integrated senescence / CAR-T / fibrosis brief](https://brownbio.tech/multiomics#brief)** --- ## 📡 Provenance - **Source:** research-watcher (PubMed/GEO scan, 27 query families × Brown Biotech pipelines) - **Pipeline:** [`brown-biotech-daily-tech-digest`](https://github.com/ohbryt/brown-biotech-platform) — 06:00 KST cron - **Generated:** 2026-07-20 06:16 KST - **Repo:** `ohbryt/brown-biotech-platform` --- _Auto-published by `brown_biotech_research_digest_publisher.py` · [brownbio.tech](https://brownbio.tech) · Decision-ready research, daily._
Brown Biotech Research Digest — 2026-07-20
PubMed/GEO scan · research-watcher · 06:00 KST