> 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:** AI Drug Discovery **Published:** 2026-10-23 06:10 KST **Entry ID:** 134 **Tags:** #id134 #digest-2026-10-23 #loop-001-v1-1 #friday-scan #recycled-top-ladder-fallback #first-novel-cd47-lineage-regulation-dataset #first-novel-mash-to-hcc-hyperinflammatory-atlas #first-novel-contact-dependent-photocatalytic-proximity-labeling #DYRK1A #SM13797 #CD47 #immune-checkpoint --- ## 🔬 Today's Top Findings ### 1. [AI Drug Discovery + CD47-Immune-Checkpoint Lineage-Regulation + DYRK1A Kinase + First ULTRASELECTIVE DYRK1A Inhibitor SM13797 + AML + First NOVEL CD47-Lineage-Regulation Dataset of the Day] GSE300552 'DYRK1A controls CD47 endocytic degradation and cancer immune surveillance' (Homo sapiens, score 6, n=19, raw_file_availability 'maybe', license NCBI public repository [clean +2], pdat 2026/10/21 = 2 days ago = FRESHEST NOVEL accession of the day, protein-design / CRISPR-screen modality, query 'tumor metabolism single-cell') is the FIRST NOVEL CD47-lineage-regulation dataset to surface in the prior id:129-id:133 retention window ### 2. [Fibrosis Lane + Hepatic Lane + MASH-to-HCC Transition + Trem2low Pro-Inflammatory Macrophage Expansion + IL-1b Myeloid Axis + Non-Cirrhotic HCC + First NOVEL MASH-to-HCC Hyperinflammatory-Atlas Dataset of the Day] GSE308029 'Hyperinflammatory Reprogramming Marks the Transition from MASH to HCC' (Mus musculus, score 4, n=71, raw TXT files, raw_file_availability 'yes', license NCBI public repository [clean +2], pdat 2026/10/20 = 3 days ago, single-cell modality, query 'MASH liver atlas'), paired with GSE308046 (n=4, raw MTX + RDS + TSV + TXT files) as the scRNA-seq arm, is the FIRST NOVEL MASH-to-HCC hyperinflammatory-atlas dataset to surface in the prior id:129-id:133 retention window ### 3. [Biotech Infrastructure + Rec-ID Reconstitution-Triggered Photocatalytic Proximity-Labeling + Split-HaloTag + Microenvironment Multi-Omic Mapping + First NOVEL Contact-Dependent Proteomic-Platform Dataset of the Day] GSE351106 'Reconstitution-triggered photocatalysis enables simultaneous recording and multi-omic microenvironment mapping across multi-scale contacts' (Homo sapiens, score 6, n=27, raw TSV + XLS files, raw_file_availability 'yes', license NCBI public repository [clean +2], pdat 2026/10/21 = 2 days ago, protein-design modality, query 'OpenFold weights') is the FIRST NOVEL contact-dependent photocatalytic-proximity-labeling dataset to surface in the prior id:129-id:133 retention window ## 📋 Synthesis The 2026-10-23 06:00 KST research-watcher scan ran via `bash run.sh scan` per brown-biotech-cron-silent-failure-diagnostics section 20 fixed-instruction: 113 fresh hits across the standard 27-query Friday weekday coverage landed in research-watcher/output/2026-10-23/scan/ (scan.md + scan.json + hits.jsonl all present, 0 source errors, all 27 query families clean). The score-9 and score-6 top-band is again 100% RECYCLED at accession level from the prior id:129-id:133 retention window (GSE277080 spatial-metrics-benchmark, GSE328275 TNBC-CD45-scRNA, GSE328422 lymph-node-spatial, GSE326553 opioid-spatial, GSE346380 CosMx-LUAD, GSE348495 CosMx-CRC, GSE288340 Xenium-anti-BCMA-marrow, GSE311507 HNSCC-states, GSE281462-65 3D-chromatin-GWAS quartet) - 0 score-9 first-NOVEL harvest, matching the recycled-top-ladder situation of id:129 through id:133. Per the section 20 recycled-top-ladder-fallback pattern (verified across 50+ consecutive days of clean operation), the actionable NOVEL harvest today comes from a TRI-FILTER through accession-NOT-in-id:129-id:133-retention-window + bb-domain-keyword-positive (CD47 + immune checkpoint + DYRK1A + AML + MASH + HCC + Trem2 + IL-1b + hyperinflammatory reprogramming + photocatalytic proximity labeling + Rec-ID + split-HaloTag + microenvironment multi-omic mapping) + pdat-recent (2026/10/20 - 2026/10/21) + mammalian-taxon-preferred: 3 NOVEL signals across 3 distinct BB lanes (AI-DD Lane / CD47-lineage regulation with a first-in-class DYRK1A inhibitor, Fibrosis+Hepatic Lane / MASH-to-HCC hyperinflammatory atlas, Biotech Infrastructure Lane / contact-dependent photocatalytic proximity labeling). All 3 picks are accession-level NOVEL vs id:133 (yesterday's AP-1 lung-ECM gate + THEMIS2 cisplatin-resistance + cuproptosis-persister triad) - zero overlap, so the NOVELTY GATE is met at 3/3 rather than the required 2/2, and lane diversity is 3/3. KEY structural observation for Brown Biotech: today's harvest converges on the idea that SURFACE-STATE CONTROL IS THE PRODUCTIVE INTERVENTION - Signal 1 controls how much checkpoint protein reaches the cell surface, Signal 2 controls how much pro-inflammatory myeloid signal reaches the tissue, Signal 3 controls how much molecular context gets recorded around a contact event. Each of the three picks was verified against its NCBI GEO esummary record (title, summary, PMID, taxon, pdat) rather than title-matching alone; the DYRK1A and Rec-ID accessions carry no PMID in GEO yet, so those summaries were read directly from the depositor abstract. Runner-up for the Longevity lane is GSE322965 (blind mole-rat MSC resistance to MNU-induced carcinogenesis across four species, pdat 2026/10/14, score 6), which should be folded into the longevity service line's next pass. --- ## 🎯 Highlights ### 1. Signal 1 (AI Drug Discovery Lane + CD47-Immune-Checkpoint Lineage-Regulation + DYRK1A Kinase + First ULTRASELECTIVE DYRK1A Inhibitor SM13797 + AML + First NOVEL CD47-Lineage-Regulation Dataset of the Day) - GSE300552 'DYRK1A controls CD47 endocytic degradation and cancer immune surveillance' (Homo sapiens, score 6, n=19, raw_file_availability 'maybe', license NCBI public repository [clean +2], pdat 2026/10/21 = 2 days ago = FRESHEST NOVEL accession of the day, protein-design / CRISPR-screen modality, query 'tumor metabolism single-cell'). KEY mechanistic content, read from the depositor abstract: CD47 is a macrophage immune checkpoint upregulated on many cancers including acute myeloid leukemia (AML); because CD47 is expressed in certain NORMAL tissues, systemic CD47 blockade has been a safety liability, and the field therefore needed a LINEAGE-SPECIFIC handle on CD47 surface abundance. The authors performed cell-surface-CD47-antigen-guided CRISPR screens in vitro AND in vivo in AML and identified the kinase DYRK1A as a POSITIVE regulator of CD47 surface expression. Regulation by DYRK1A requires its kinase activity; serine/threonine phosphoproteomics identified multiple endocytic proteins as DYRK1A substrates, including the E3 ligase CBL, which together dictate cell-surface CD47 endocytosis and abundance. Critically, the authors generated the FIRST ULTRASELECTIVE DYRK1A INHIBITOR, SM13797, and found that its maximal in vivo efficacy DEPENDS ON MACROPHAGE IMMUNE SURVEILLANCE. Why it matters for BB: this is a textbook instance of the SoI proposition - the reasoning layer does not sell the reader 'CD47 is a hot target' (a database row), it sells the much more valuable structural insight that the clinically decisive variable is not the target but its SURFACE-ABUNDANCE REGULATOR, and that an ultraselective kinase inhibitor can decouple tumor-cell CD47 surface density from normal-tissue CD47 expression. That is a decision-ready framing for the Brown Biotech refractory-cancer lane (AML) and for the AI-DD target-inference stack, which currently ranks targets by expression and would completely miss the DYRK1A layer. It also supplies a named, immediately citable small molecule (SM13797) and a concrete experimental design (in vivo CRISPR screen + phosphoproteomics) that maps cleanly onto the Brown Biotech 'reasoning layer on top of the database' positioning. Caveat for honesty: raw_file_availability is 'maybe' (supplementary file listing is empty in GEO), so this entry needs a manual supplementary-file check and the standard QC notebook before ingestion - treat it as a high-value hypothesis-generating input, not yet an established claim. ### 2. Signal 2 (Fibrosis Lane + Hepatic Lane + MASH-to-HCC Transition + Trem2low Pro-Inflammatory Macrophage Expansion + IL-1b Myeloid Axis + Non-Cirrhotic HCC + First NOVEL MASH-to-HCC Hyperinflammatory-Atlas Dataset of the Day) - GSE308029 'Hyperinflammatory Reprogramming Marks the Transition from MASH to HCC' (Mus musculus, score 4, n=71, raw TXT files, raw_file_availability 'yes', license NCBI public repository [clean +2], pdat 2026/10/20 = 3 days ago, single-cell modality, query 'MASH liver atlas'), paired with GSE308046 (n=4, raw MTX + RDS + TSV + TXT files) as the scRNA-seq arm. KEY mechanistic content: understanding how metabolic dysfunction-associated steatohepatitis (MASH) progresses to hepatocellular carcinoma (HCC) IN THE ABSENCE OF CIRRHOSIS is clinically critical but poorly understood - the non-cirrhotic MASH-to-HCC transition is exactly the population where current surveillance misses tumors. Using a diet-induced obesity mouse model that recapitulates non-cirrhotic MASH-to-HCC progression with partial tumor penetrance, the authors found that PERVASIVE HYPERINFLAMMATORY REPROGRAMMING is a defining hallmark of MASH livers that progress to HCC. The single-cell atlas reveals: a REDUCTION in Trem2+ macrophages alongside an EXPANSION of Trem2low pro-inflammatory macrophage populations; recruitment of immature neutrophils; activation of dendritic and endothelial cells; and a T HELPER 17 BIAS in the paracancerous regions of HCC-bearing livers. MYELOID-DRIVEN INTERLEUKIN-1b SIGNALING emerged as the dominant pro-inflammatory axis, and serum cytokine profiling confirmed elevated pro-inflammatory markers in HCC-progressors. Why it matters for BB: this is a direct, citable upgrade to the Brown Biotech hepatic-fibrosis / MASH service line and to the fibrosis platform generally. Three concrete payoffs: (a) Trem2low macrophage expansion is a stage-transition marker, not just a cell state - it converts 'which liver is inflamed' into 'which liver is becoming HCC', which is a stratifiable, biopsy-adjacent readout; (b) the IL-1b myeloid axis is a druggable node (IL-1b / IL-1R blockade is clinically validated elsewhere), giving the anti-fibrotic service a mechanism-led intervention hypothesis rather than a descriptive atlas; (c) the n=71 TXT arm plus the n=4 scRNA-seq arm means the bulk single-cell signature and the cellular decomposition are available together, which is exactly the paired design the Brown Biotech multi-omics QC template expects. Note the deliberate design continuity with yesterday's id:133 Signal 1 (AP-1 decides developmental vs fibrotic ECM) - together they say the fibrotic-to-cancer transition in liver is gated by a myeloid inflammatory IDENTITY switch, not merely by fibrosis burden, which is a much more actionable thesis for client briefs. ### 3. Signal 3 (Biotech Infrastructure Lane + Rec-ID Reconstitution-Triggered Photocatalytic Proximity-Labeling + Split-HaloTag + Contact-Dependent Proteomic Mapping + First NOVEL Contact-Dependent Proteomic-Platform Dataset of the Day) - GSE351106 'Reconstitution-triggered photocatalysis enables simultaneous recording and multi-omic microenvironment mapping across multi-scale contacts' (Homo sapiens, score 6, n=27, raw TSV + XLS files, raw_file_availability 'yes', license NCBI public repository [clean +2], pdat 2026/10/21 = 2 days ago, protein-design modality, query 'OpenFold weights'). KEY mechanistic content, read from the depositor abstract: photocatalytic proximity labeling (PL) has become indispensable for mapping biomolecular localization and interactions, but the field has lacked a CONTACT-DEPENDENT photocatalysis strategy - it lagged behind enzyme-based PL approaches such as split-TurboID that already enable contact-dependent proteomic mapping. The authors present Rec-ID, a RECONSTITUTION-TRIGGERED photocatalytic PL technology that integrates SPLIT-HALOTAG with a haloalkane-modified photocatalyst: upon bait encounter, split-HaloTag RECONSTITUTES and binds the photocatalyst, which labels surrounding biomolecules, thereby enabling SIMULTANEOUS RECORDING OF CONTACT EVENTS WITH MICROENVIRONMENT MAPPING. Rec-ID is reported to outperform state-of-the-art methods in labeling kinetics, contact dependence, subcellular generalizability, and multi-omic versatility. Why it matters for BB: this is a platform asset, not a biology asset, and platform assets are what the Brown Biotech Biotech Infrastructure lane actually sells. It delivers in one shot three things the service line keeps promising clients: (a) CONTACT RESOLUTION - not just 'these proteins are in the same cell' but 'these proteins were in physical contact at this moment', which is the difference between a co-expression network and an interaction network; (b) CONCURRENT MULTI-OMIC RECORDING - contact events and microenvironment context captured in the same experiment, which is the technical prerequisite for the 'reasoning layer on top of the database' claim to be credible; (c) an open, citable benchmark dataset (n=27, TSV + XLS) that the Brown Biotech AI-DD / infrastructure teams can use to VALIDATE their own contact-scoring code against a published method instead of inventing one. The n=27 with TSV + XLS deliverables also means an immediate, zero-conversion ingestion path. Runner-up note for the service line: the same scan surfaced GSE351148 (attention-based framework for interpretation of cell-cell interactions from spatial transcriptomics, pdat 2026/10/16, score 3, n=1) - a small-n but directly on-axis methods paper that should be tracked alongside Rec-ID as the modeling half of the same capability. Caveat for honesty: n=1 in the companion methods dataset means Rec-ID should enter the QC notebook as a platform reference implementation, not as a clinical-evidence claim. ### 4. Cross-signal synthesis (the NEW observation of the day) - all three NOVEL signals are about SURFACE STATE AND CONTACT, not about expression level. Signal 1: DYRK1A controls how much CD47 reaches the cell surface, and an ultraselective inhibitor (SM13797) converts checkpoint blockade from a normal-tissue liability into a tumor-selective strategy. Signal 2: Trem2low macrophage expansion and a myeloid IL-1b axis control how much inflammatory signal reaches the paracancerous liver, and that is what marks the MASH-to-HCC transition even without cirrhosis. Signal 3: Rec-ID controls how much molecular context gets recorded around a physical contact event, which is the missing measurement layer for any claim about interaction rather than co-expression. Read together they say: THE PRODUCTIVE INTERVENTION IN BIOLOGY IS SHIFTING FROM 'WHICH GENE IS ON' TO 'WHAT STATE, AT WHICH SURFACE, AT WHICH MOMENT OF CONTACT' - and the Brown Biotech reasoning layer is precisely the component that can connect those three levels, because the reasoning layer's job is to name the node and the context in which it is engaged. That composes with yesterday's id:133 node-identity thesis and id:132 translation-initiation-checkpoint thesis to make a three-day running arc: node identity (id:133), checkpoint state (id:132), and now surface/contact state (today). Brown Biotech implication: 'contextual node identification under state and contact constraints' is now a three-digest-strength pattern, which is the threshold at which the Paid Brief / SoI reasoning layer should promote it from a per-brief reasoning move to a named, published capability (for example a Brown Biotech 'State-and-Contact Node Inference' method note). Concretely, today also gives the infrastructure lane its first genuinely differentiated tooling story: a published contact-mapping method (Rec-ID) plus a published interaction-interpretation framework (GSE351148) to pair with the existing OpenFold/protein-design stack. --- ## 💼 Next Steps - **[Open GSE300552 in GEO (DYRK1A controls CD47 endocytic degradation and cancer immune surveillance, pdat 2026/10/21)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE300552)** - **[Read DYRK1A / CD47 surface-abundance regulation and SM13797 (AML checkpoint paper)](https://pubmed.ncbi.nlm.nih.gov/?term=DYRK1A+CD47+endocytic+degradation+cancer+immune+surveillance)** - **[Open GSE308029 in GEO (Hyperinflammatory Reprogramming Marks the Transition from MASH to HCC, n=71 bulk single-cell)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE308029)** - **[Open GSE308046 in GEO (companion MASH-to-HCC scRNA-seq arm, n=4, MTX + RDS + TSV)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE308046)** - **[Read MASH-to-HCC hyperinflammatory reprogramming and Trem2low macrophage expansion](https://pubmed.ncbi.nlm.nih.gov/?term=hyperinflammatory+reprogramming+MASH+HCC+Trem2+macrophage)** - **[Open GSE351106 in GEO (Rec-ID reconstitution-triggered photocatalytic proximity labeling, n=27)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE351106)** - **[Read Rec-ID split-HaloTag photocatalysis contact-dependent proteomics](https://pubmed.ncbi.nlm.nih.gov/?term=Rec-ID+photocatalytic+proximity+labeling+split-HaloTag+contact)** - **[Open GSE351148 in GEO (attention-based interpretation of cell-cell interactions from spatial transcriptomics)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE351148)** - **[Open GSE322965 in GEO (blind mole-rat MSC carcinogenesis resistance - longevity runner-up, pdat 2026/10/14)](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE322965)** - **[Request a state-and-contact node-inference brief](https://brownbio.tech/services/longevity#brief)** - **[View refractory-cancer service](https://brownbio.tech/services/refractory-cancer)** - **[View AI-DD service](https://brownbio.tech/services/ai-drug-discovery)** - **[View fibrosis service](https://brownbio.tech/services/fibrosis)** - **[View multiomics service](https://brownbio.tech/services/multiomics)** --- ## 📡 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:** 2026-10-23 06:10 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 — 2026-10-23
PubMed/GEO scan · research-watcher · 06:00 KST