Research Pulse·2026-06-07

Research Pulse — 2026-06-07

28 journals × 7 topics · fibrosis · OXPHOS · ferroptosis · sarcopenia · senescence

# Research Pulse — 2026-06-07
_Generated 2026-06-07 07:58 KST_

Daily Brown Biotech research pulse. PubMed/PMC scan across 28 journals × 7 topics,
grouped into query families aligned with active Brown Biotech pipelines.

### 1. OXPHOS & Mitochondrial Vulnerability (0 hits)
_Brown Biotech relevance: KEAP1/NRF2, MASH lipotoxicity, ARP v27 target #18_

- (no fresh hits today)

### 2. MASH Liver Atlas & DGAT2 (0 hits)
_Brown Biotech relevance: DGAT1/2 TPP, Sargassum japonica pipeline_

- (no fresh hits today)

### 3. Ferroptosis & Lipid Peroxidation (0 hits)
_Brown Biotech relevance: GPX4, ACSL4, lipid droplet biology, KEAP1 mechanism_

- (no fresh hits today)

### 4. Sarcopenia Single-Cell (0 hits)
_Brown Biotech relevance: D+Q + pioglitazone senolytic, HBAT brown adipose_

- (no fresh hits today)

### 5. Aging Muscle Atlas (0 hits)
_Brown Biotech relevance: longevity-db cc-by-4.0, cross-tissue aging clock_

- (no fresh hits today)

### 6. Frailty Multi-Omics (2 hits)
_Brown Biotech relevance: longevity platform, anti-aging cosmetics research_

- [House com­mit­tee di­rects FDA to speed up IND sign-off process](https://endpoints.news/house-committee-directs-fda-to-speed-up-ind-sign-off-process/) — Endpoints News
- [FDA recall: Atomoxetine Capsules HCL, 10 mg Capsules in unit dose foil strip, strips are packed in cartons of 100, Rx Only, Pkg by:](https://www.fda.gov/safety/recalls-market-withdrawals-safety-alerts#D-0538-2026) — FDA

### 7. Muscle Fibrosis Spatial Transcriptomics (0 hits)
_Brown Biotech relevance: anti-fibrotic pipeline, FAP subsets_

- (no fresh hits today)

### 8. Skeletal Muscle Aging Visium (0 hits)
_Brown Biotech relevance: spatial workflow validation, type II fiber atrophy_

- (no fresh hits today)

### 9. Human Aging Cohort Metabolomics (4 hits)
_Brown Biotech relevance: longevity diagnostics, biomarker discovery_

- [ADC biotech Ona Ther­a­peu­tics se­cures $86.6M in Se­ries B raise](https://endpoints.news/adc-biotech-ona-therapeutics-secures-86-6m-in-series-b-raise/) — Endpoints News
- [Isomorphic Labs announces Series B investment round](https://www.isomorphiclabs.com/articles/isomorphic-labs-announces-series-b-investment-round) — Isomorphic Labs
- [Careers Inside Isomorphic Labs' Lausanne office 18.02.25](https://www.isomorphiclabs.com/articles/inside-isomorphic-labs-lausanne-office) — Isomorphic Labs

### 10. Fibroblast Atlas (0 hits)
_Brown Biotech relevance: CTHRC1+ fibrogenic subset, CMap repurposing_

- (no fresh hits today)

### 11. FAP Spatial Transcriptomics (0 hits)
_Brown Biotech relevance: cross-tissue FAP atlas, IPF lung_

- (no fresh hits today)

### 12. MASH Liver Atlas (snRNA-seq) (0 hits)
_Brown Biotech relevance: hepatic stellate cell activation trajectories_

- (no fresh hits today)

### 13. Lung Fibrosis Spatial Atlas (0 hits)
_Brown Biotech relevance: IPF CTHRC1+ signaling hub, bleomycin model_

- (no fresh hits today)

### 14. Senescence & Senolytic (0 hits)
_Brown Biotech relevance: D+Q + pioglitazone, HBAT activation, senotype deep learning_

- (no fresh hits today)

### Other live signals (2 hits)
- [Cullinan's new data makes clearest case yet for autoimmune TCEs](https://www.fiercebiotech.com/biotech/it-looks-hypothesis-will-pan-out-cullinan-builds-case-autoimmune-tces) — Fierce Biotech
- [In Wash­ing­ton, biotech's Chi­na hawks are win­ning the con­ver­sa­tion](https://endpoints.news/in-washington-biotechs-china-hawks-are-winning-the-conversation/) — Endpoints News

---

**Totals:** 6 query-family hits across 14 families.

_Source: Brown Biotech daily signal feed (`brown_biotech_daily_signal_feed.py`). For deep-dive synthesis, see `/Users/ocm/openclaw/workspace/arp-v27/literature/`._

---

## 🔬 Featured Reference: scGPT (Nature Methods 2024)

**Added 2026-06-07** — the canonical single-cell foundation model for this research pulse.

- **Title:** scGPT: toward building a foundation model for single-cell multi-omics using generative AI
- **Authors:** Cui H, Wang C, Maan H, Pang K, Luo F, Duan N, Wang B (Bo Wang lab, U Toronto / Vector Institute)
- **DOI:** 10.1038/s41592-024-02201-0 | Nature Methods Vol 21, 1470–1480 (Aug 2024)
- **Code:** https://github.com/bowang-lab
- **PDF:** `/Users/ocm/.hermes/cache/documents/doc_78d3553e9224_s41592-024-02201-0.pdf`

### Why this paper matters for Brown Biotech

scGPT is a generative pretrained transformer trained on **33 million human cells** (CELLxGENE, 51 organs, 441 studies). It enables 5 downstream tasks: cell type annotation, multi-batch integration, multi-omic integration, **perturbation response prediction**, and gene regulatory network inference.

**Maps to 5 of 14 query families in this research pulse:**
- **Sarcopenia Single-Cell** — direct (cell-type annotation in aged muscle)
- **Aging Muscle Atlas** — direct (cross-tissue aging clock)
- **Fibroblast Atlas** — direct (CTHRC1+ fibrogenic subset annotation)
- **MASH Liver Atlas (snRNA-seq)** — direct (multi-omic integration)
- **Senescence & Senolytic** — indirect (perturbation prediction for senolytic combination)

### Key benchmark results
- **Cell type annotation:** Outperforms TOSICA, scBERT (hPancreas, ~0.85 accuracy)
- **Perturbation prediction:** Outperforms GEARS by 5-20% margins (Adamson, Norman, Replogle datasets)
- **Multi-omic integration:** Only method that successfully integrates RNA + ATAC + protein with batch correction
- **Cross-tissue transfer:** PBMC 10k works **without fine-tuning** (highlights pretraining generalizability)

### Brown Biotech integration

| Track | Action |
|---|---|
| **ARP v27** | Adopt pretrain→fine-tune paradigm for Foundation Model engine. Use for Sargassum japonica perturbation screens. |
| **Paid Briefs** | New SKU: "Single-Cell Cell-Type Annotation" ₩2M~5M |
| **biostatx** | Default cell-type annotation backend (replaces Seurat-only); multi-omic integration for FFPE Xenium |
| **PRISM RAG** | ✓ Ingested 2026-06-07, 85 chunks, FAISS index 659 → 744 vectors |

**Deep-dive analysis:** `/Users/ocm/openclaw/workspace/arp-v27/literature/scGPT_Deep_Analysis.md` (9.4KB, 9 sections)

---

## 🔬 Featured Reference: turbovec + TurboQuant (2026)

**Added 2026-06-07** — drop-in vector index for PRISM RAG and Brown Biotech services.

- **Repo:** https://github.com/RyanCodrai/turbovec (5,597 stars, MIT)
- **Paper:** arXiv 2504.19874 (Google Research TurboQuant algorithm)
- **PyPI:** `pip install turbovec` (v0.7.0 installed)
- **PRISM backend:** `/Volumes/4TB/prism/rag/turbovec_store.py` (new, drop-in parallel to FAISS)

### Why this matters for Brown Biotech

turbovec is a Rust + Python vector index built on Google Research's **TurboQuant** — a data-oblivious quantizer that matches the Shannon lower bound on distortion. **No codebook training, no separate train phase.** Hand-written NEON (ARM) and AVX-512 (x86) SIMD kernels.

**Smoke test verified 2026-06-07 on PRISM 744 chunks:**

| Metric | FAISS (TF-IDF) | turbovec (MiniLM 4-bit) | Delta |
|---|---|---|---|
| **Disk** | 15.29 MB | **0.15 MB** | **103.5× smaller** |
| **Search latency** | ~ms | **0.03 ms/query** | ~30× faster |
| **Add time (744 vec)** | rebuild needed | 0.05s (14,871 vec/s) | **Online** |
| **Filtered search** | post-filter (recall hit) | **native kernel allowlist** | New capability |
| **Semantic matching** | lexical only (TF-IDF) | MiniLM 384-dim dense | Cross-family terms work |

### Brown Biotech integration paths

| Service | Use | Action |
|---|---|---|
| **PRISM** | Stage 1: parallel to FAISS. Stage 3: 35M abstracts | Drop-in backend done |
| **biostatx** | LangChain client RAG | Swap `InMemoryVectorStore` → `turbovec.langchain.TurboQuantVectorStore` |
| **brown_biotech publisher** | Semantic query family mapping | Replace keyword `map_to_family()` with MiniLM classifier |
| **Inventa** | (If RAG needed) | LlamaIndex drop-in |

**Multi-tenant RAG:** `IdMapIndex.search(allowlist=tenant_ids)` is native to turbovec. 100s of clients in one process, isolated by ID range. **Currently impossible** with FAISS without separate indexes per tenant.

**Deep-dive:** `/Users/ocm/openclaw/workspace/arp-v27/literature/turbovec_TurboQuant_Analysis.md` (12 KB, 11 sections)
**Skill:** `~/.hermes/skills/brown-biotech-turbovec-prism-integration/SKILL.md`
**Notion:** https://www.notion.so/378f273533a4818ca9aec955c565b7ad

Raw data + scoring: Daily Tech Digest (separate feed)

Request a Paid Brief