Research Pulse — 2026-09-17 | Brown Biotech

# Research Pulse — 2026-09-17
_Generated 2026-09-17 07:30 KST · AI × Bioresearch Daily synthesis_

Brown Biotech daily research pulse — multi-front synthesis across CRISPR /
gene editing, AI drug discovery, protein AI, single-cell foundation models,
FDA/regulatory, and spatial transcriptomics. Sources: company filings,
peer-reviewed journals, FDA guidance, conference disclosures.

---

### Post 1: 오늘의 전체 흐름

🧬 **AI × Bioresearch Daily — 9월 17일 (목요일) · Scaling(확산) Day**

9/17(확산) → 9/16(종합) → 9/15(검증) → 9/14(정밀화) → 9/13(가동화) → 9/12(다양화) → 9/11(전이) → 9/10(재보정). Synthesizing(9/16)된 통합 프레임워크들이 **대규모 임상 적용·산업 확산·규제 표준화(Scaling) 단계로 이전하며 실제 파이프라인规模和 임상 데이터로 확산되는 날** — Lushington et al.으로 PE first-in-human functional restoration + Yu et al.으로 PE7·PE8 architecture evolution + Mengistu et al.으로 CRISPR-Cas临床 applications comprehensive review(phase I-III trial outcomes), Jia et al.으로 DrugCLIP contrastive learning genome-wide virtual screening — docking 대비 10M× 속도 + GenomeScreenDB 10,000 proteins × 500M compounds(Science 2026), Singh et al.으로 AlphaFold-like models drug discovery에 protein-ligand data 필요성 체계적 분석(Current Opinion Structural Biology 2026), Shahzad et al.으로 FDA 38 CGTs 86.8% orphan + 92.1% expedited + 73.7% 단일 pivotal study + postmarketing primary efficacy 17.9%(Clinical Pharmacology Therapeutics 2026), Li et al.으로 12개 암종 56 LCPs + Niche_4 poor prognosis + Niche_11 better survival(Cell Reports Medicine 2026)이 동시에 진행된다.

Scaling은 Synthesizing의 확산 단계다. 통합 프레임워크들이 실제 임상 파이프라인과 대규모 데이터로 확산되며 산업적 적용 가능성과 규제 표준화를 동시에 확인하는 날이다.

*출처: [Lushington PE First-in-Human Molecular Therapy 2026](https://consensus.app/papers/details/79a173589b1d58598fd94fa334768881/), [Yu PE Evolution Advanced Science 2026](https://consensus.app/papers/details/475fc3cbbe805f5e915626f08d2da039/), [Mengistu CRISPR Cas Systems Gene Therapy Journal of Gene Medicine 2026](https://consensus.app/papers/details/6f6486cb3f3f5baab6f2ebaf3e67c39d/), [Jia DrugCLIP Genome-Wide Virtual Screening Science 2026](https://consensus.app/papers/details/5acea8f3221b5acfa37184d1fe50da28/), [Singh AlphaFold Protein Ligand Drug Discovery Current Opinion Structural Biology 2026](https://consensus.app/papers/details/81a3128ae1615cfeb1bd6544d3ad2f1b/), [Shahzad FDA CGT Assessment Clinical Pharmacology Therapeutics 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Li Pan-cancer Spatial Transcriptomics Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/)*

*#ScalingDay #Lushington #MolecularTherapy #PrimeEditing #FirstInHuman #Yu #AdvancedScience #PE7 #PE8 #Mengistu #JournalOfGeneMedicine #CRISPRCas #ClinicalTrials #Jia #DrugCLIP #Science #ContrastiveLearning #VirtualScreening #10MillionTimes #Faster #Docking #GenomeScreenDB #Singh #CurrentOpinion #StructuralBiology #AlphaFold #ProteinLigand #DataNeeded #Shahzad #ClinicalPharmacologyTherapeutics #FDA #CGTs #OrphanDesignation #ExpeditedPathway #Postmarketing #Li #CellReportsMedicine #PanCancer #LCP #Niche4 #Niche11 #Synthesizing #Scaling #ClinicalTranslation #BiomedicalAI #DailyReport*

---

### Post 2: CRISPR & Gene Editing — PE First-in-Human + PE7/PE8 Architecture Evolution과 Clinical Trials Phase I-III Outcomes

🪄 **[CRISPR/Gene Editing] Lushington et al.으로 prime editing first-in-human functional restoration + promising safety profile(Molecular Therapy 2026) + Yu et al.으로 PE1→PE7 evolution + PE7 architecture innovations + efficiency/capacity improvements(Advanced Science 2026) + Mengistu et al.으로 CRISPR-Cas systems gene therapy comprehensive review — base editing·prime editing·epigenome modulation + Phase I-III trial outcomes + off-target·delivery·immunogenicity challenges(Journal of Gene Medicine 2026): CRISPR Scaling(확산) 동시 확인**

CRISPR 유전자 편집 기술의**대규모 임상 적용과 기술 확산(Scaling)**이 출처에서 동시에 확인된다[1][2][3].

**PE First-in-Human: Functional Restoration + Promising Safety Profile[1]**
C. Lushington et al.(Molecular Therapy 2026)의 review는 prime editing의 first-in-human study 결과를 분석한다. **PE 개요**: Cas9 nickase + engineered reverse transcriptase(RT) + pegRNAfusion — DNA 이중가닥 절단 없이 정밀 편집. **핵심 성과 1 — Functional restoration 확인**: 첫 인체 실험에서 기능적 회복 확인. **핵심 성과 2 — Promising safety profile**:迄今为止 안전한 프로파일. **핵심 성과 3 — pegRNA improvements**: PE 효율성 향상을 위한 pegRNA 개선. **핵심 성과 4 — Delivery advances**: nanoparticles와 split viral systems — 전임상 질환 모델에서 적용 확대. **핵심**: PE first-in-human functional restoration — CRISPR Scaling Consolidating.

**PE Evolution: PE1→PE7 Architecture Innovations[2]**
J. Yu et al.(Advanced Science 2026)의 review는 prime editing의 기술 진화를 종합한다. **PE1→PE7 진화**: (i) Cas variant selection 개선, (ii) RT engineering 진전, (iii) pegRNA improvements. **핵심 성과 1 — Editing efficiency 향상**: 각 세대별 효율성 진전. **핵심 성과 2 — Editable target range 확대**: 더 넓은 표적 범위. **핵심 성과 3 — Large genomic region correction 가능**: 대용량 유전체 교정 가능. **핵심 성과 4 — PE7 architecture innovations**: 최신 아키텍처 혁신들. **적용 분야**: inherited disorders, cancer, infectious diseases, neurodegenerative diseases. **핵심**: PE1→PE7 evolution + architecture innovations — CRISPR Scaling Consolidating.

**CRISPR-Cas Systems Gene Therapy: Phase I-III Trial Outcomes + Challenges[3]**
G. Mengistu et al.(The Journal of Gene Medicine 2026)의 review는 CRISPR-Cas systems의 임상 적용을 종합한다. **기술 범위**: CRISPR nucleases + base editing + prime editing + epigenome modulation. **핵심 성과 1 — Phase I-III trial outcomes**: 조기 임상 시험 결과 종합. **핵심 성과 2 — Ex vivo/in vivo 모두 적용**: 두 접근 모두 임상 가능 확인. **핵심 성과 3 — Off-target activity**: 여전히 주요 도전과제. **핵심 성과 4 — Delivery inefficiency**: 특히 in vivo 적용에서. **핵심 성과 5 — Immune responses to Cas proteins**: 면역 반응 문제. **핵심 성과 6 — Editing heterogeneity**: 편집 이질성. **핵심 성과 7 — Long-term safety uncertainties**: 장기 안전성 불확실성. **핵심**: CRISPR-Cas systems Phase I-III outcomes + remaining challenges — CRISPR Scaling Consolidating.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Sawtell macrocyclic covalent peptide**로 **새로운 치료 플랫폼**(Synthesizing)을, **Na IS110 dual-match**로 **새로운 gene delivery 메커니즘**(Synthesizing)을, **Ambardekar ACC cardiovascular**로 **심혈관 질환 적용**(Synthesizing)을 제시했다면, 오늘은 **Lushington PE first-in-human**으로 **실제 인체 실험 기능적 회복 + 안전한 프로파일**(Scaling — 임상 이행里程碑)을(new angle), **Yu PE7/PE8 evolution**으로 **architecture innovations를 통한 효율성 향상**(Scaling — 기술 확산)을(new angle), **Mengistu Phase I-III outcomes**로 **CRISPR-Cas 임상 시험 전체 스펙트럼 종합**(Scaling — 산업 확산)을 동시에 제시한다. Synthesizing(9/16) → CRISPR Scaling(9/17).

실무 함의: 유전자 치료 임상 개발팀에서 Lushington et al.의 PE first-in-human 결과를 참조하여 prime editing clinical trial design strategy를 수립해야 한다.

*출처: [Lushington PE First-in-Human Molecular Therapy 2026](https://consensus.app/papers/details/79a173589b1d58598fd94fa334768881/), [Yu PE Evolution Advanced Science 2026](https://consensus.app/papers/details/475fc3cbbe805f5e915626f08d2da039/), [Mengistu CRISPR Cas Systems Gene Therapy Journal of Gene Medicine 2026](https://consensus.app/papers/details/6f6486cb3f3f5baab6f2ebaf3e67c39d/)*

*#CRISPR #GeneEditing #Lushington #MolecularTherapy #PrimeEditing #FirstInHuman #FunctionalRestoration #SafetyProfile #Yu #AdvancedScience #PE1 #PE7 #PE8 #Architecture #Evolution #Efficiency #Capacity #Mengistu #JournalOfGeneMedicine #CRISPRCas #BaseEditing #EpigenomeModulation #PhaseI #PhaseII #PhaseIII #ClinicalTrials #OffTarget #Delivery #Immunogenicity #LongTermSafety #ScalingDay #ClinicalTranslation #IndustrialScaling #BiomedicalAI #DailyReport*

---

### Post 3: AI 신약 — DrugCLIP 10M× Virtual Screening + GenomeScreenDB 10,000×500M Compounds와 AlphaFold-Ligand Data Gap

💊 **[AI 신약] Jia et al.으로 DrugCLIP contrastive learning genome-wide virtual screening — docking 대비 10,000,000× 속도 + AlphaFold2 predicted structures로 thyroid hormone receptor 17.5% hit rate + GenomeScreenDB 10,000 proteins × 500M compounds 사전 계산DB 공개(Science 2026) + Singh et al.으로 AlphaFold-like models drug discovery에 protein-ligand data 필요성 체계적 분석 — PDB 50년 축적에도 불구하고 실제 약물 발견 태스크를 위한 dataset 확대 필요(Current Opinion Structural Biology 2026) + Mao et al.으로 AI Phase I 80-90%(Hist. 52%) BUT Phase II ~40% 근본 원인 target hypothesis validity 미개선(Frontiers Pharmacology 2026): AI drug discovery Scaling(확산) 동시 확인**

AI-Driven Drug Discovery의**초고속 가상 선별과 대규모 DB 확산(Scaling)**이 출처에서 동시에 확인된다[4][5][6].

**DrugCLIP: 10M× Faster Genome-Wide Virtual Screening + GenomeScreenDB[4]**
Y. Jia et al.(Science 2026)의 연구는 DrugCLIP contrastive learning framework를 제시한다. **핵심 혁신 — 10,000,000× 속도 향상**: 기존 docking 대비. **핵심 성과 1 — Contrastive learning 적용**: protein-ligand interaction learning. **핵심 성과 2 — AlphaFold2 predicted structures 활용**: 구조 기반 가상 선별. **핵심 성과 3 — Thyroid hormone receptor 17.5% hit rate**: AlphaFold2 예측 구조만으로. **핵심 성과 4 — GenomeScreenDB 공개**: ~10,000 human proteins × 500M compounds 사전 계산 결과 — post-AlphaFold era drug discovery paradigm. **핵심 성과 5 — Wet-lab validation**: 15% hit rate (norepinephrine transporter), 17.5% hit rate (thyroid hormone receptor). **핵심**: DrugCLIP 10M× 속도 + GenomeScreenDB 공개 — AI drug discovery Scaling Consolidating.

**AlphaFold-Like Models: Protein-Ligand Data Gap for Drug Discovery[5]**
S. Singh et al.(Current Opinion in Structural Biology 2026)의 perspective는 AlphaFold-like models의 drug discovery 적용 한계를 분석한다. **문제의식**: PSPMs(Protein Structure Prediction Models)는 구조 예측은 혁신했으나 약물 발견에는 제한. **핵심 성과 1 — PDB 50년 축적**: 50년 이상의 단백질-리간드 데이터. **핵심 성과 2 — Dataset 확대 필요**: 실제 drug design 태스크를 위한 체계적 dataset 확대 필요. **핵심 성과 3 — Hierarchical framework 제안**: PSPMs의 강점과 한계를 체계적으로 분류. **핵심 성과 4 — Experimental + physics-based datasets 필요**: challenging tasks를 위한 체계적 데이터셋 생성 노력 강조. **핵심**: AlphaFold-like models protein-ligand data gap — AI drug discovery Scaling limitations.

**AI Phase I vs Phase II: Target Validity Unresolved[6]**
Z. Mao et al.(Frontiers in Pharmacology 2026)의 review는 AI drug discovery의 algorithmic foundations에서 clinical translation까지 분석한다[6]. **Phase I 성공률 80-90%**: AI-native pipelines vs historical ~52% — molecular safety engineering과 ADMET prediction 진정한 우위. **Phase II 성공률 ~40%**: AI ≈ conventional ~37% — 통계적으로 유의미한 차이 없음. **근본 원인**: Target hypothesis validity — AI는 올바른 biological premise 선택 확률 미개선. **핵심**: Phase I 우위 있으나 Phase II 차이 없음 — target validity 미개선 — AI drug discovery Scaling phase separation.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Cai multimodal AI/digital twins/closed-loop**로 **향후 통합 프레임워크**(Synthesizing)를 제시했다면, 오늘은 **Jia Science DrugCLIP**으로 **10M× 속도 향상 + GenomeScreenDB 10,000×500M大规模DB 공개**(Scaling — 산업 인프라 확산)를(new angle), **Singh Current Opinion**으로 **AlphaFold-ligand data gap으로 인한 실제 약물 발견 한계**(Scaling — 한계 인식)를(new angle), **Mao Phase I/II 분기**를 **Scaling 문맥에서 phase separation 현실**로 재제시한다. Synthesizing(9/16) → AI drug discovery Scaling(9/17).

실무 함의: 약물 발견팀에서 Jia et al.의 GenomeScreenDB를 참조하여 genome-wide virtual screening strategy를 수립해야 한다.

*출처: [Jia DrugCLIP Genome-Wide Virtual Screening Science 2026](https://consensus.app/papers/details/5acea8f3221b5acfa37184d1fe50da28/), [Singh AlphaFold Protein Ligand Drug Discovery Current Opinion Structural Biology 2026](https://consensus.app/papers/details/81a3128ae1615cfeb1bd6544d3ad2f1b/), [Mao AI Drug Discovery Frontiers Pharmacology 2026](https://consensus.app/papers/details/ca5321af0bb656e095b6108766231764/)*

*#AI #DrugDiscovery #Jia #DrugCLIP #Science #ContrastiveLearning #VirtualScreening #10MillionTimes #Faster #Docking #GenomeScreenDB #10 ThousandProteins #500MCompounds #AlphaFold2 #ThyroidHormone #17Point5Percent #HitRate #WetLab #Validation #Singh #CurrentOpinion #StructuralBiology #AlphaFold #ProteinLigand #DataGap #PDB #Mao #FrontiersPharmacology #PhaseI #PhaseII #TargetValidity #ScalingDay #VirtualScreening #LargeScale #Infrastructure #BiomedicalAI #DailyReport*

---

### Post 4: 단백질 AI — Protein-Ligand Data Gap과 AlphaFold Drug Discovery Application Limits

🔬 **[단백질 AI] Singh et al.으로 AlphaFold-like models drug discovery에 protein-ligand data 필요성 — PSPMs drug discovery 강점/한계 hierarchical framework + experimental/physics-based datasets 필요성(Current Opinion Structural Biology 2026) + Yin et al.으로 AlphaFold/protein language models comprehensive review — ESMFold/RoseTTAFold/AlphaFold3 비교 + drug discovery/enzyme engineering applications(Frontiers Molecular Biosciences 2026) + Feldman et al.으로 AlphaFold3 adversarial 35% confidence reliability 한계(Computational Structural Biotechnology 2026): 단백질 AI의 Scaling(확산) 한계 동시 확인**

단백질 AI의**약물 발견 적용 한계와 데이터 필요성(Scaling)**이 출처에서 동시에 확인된다[5][7][8].

**AlphaFold-Like Models: Protein-Ligand Data Gap[5]**
S. Singh et al.(Current Opinion in Structural Biology 2026)의 perspective는 PSPMs의 drug discovery 적용을 분석한다[5]. **핵심 성과 1 — Hierarchical framework**: PSPMs의 drug discovery 강점/한계 체계적 분류. **핵심 성과 2 — PDB 50년 축적의 한계**: 50년 데이터에도 실제 약물 발견 task를 위한 dataset 확대 필요. **핵심 성과 3 — Experimental data 필요**: challenging tasks를 위한 체계적 experimental + physics-based datasets 생성. **핵심 성과 4 — Frontier models 개발 필요**: 현재模型的 한계를 극복하기 위한 체계적 노력. **핵심**: AlphaFold-like models protein-ligand data gap — 단백질 AI Scaling limitations.

**AlphaFold Comprehensive Review: ESMFold/RoseTTAFold/AlphaFold3[7]**
T. Yin et al.(Frontiers in Molecular Biosciences 2026)의 review는 AI-driven protein structure prediction을 종합한다[7]. **핵심 성과 1 — 모델 비교**: (i) AlphaFold3: 최고 정확도, (ii) RoseTTAFold: end-to-end neural + evolutionary + geometric constraints, (iii) ESMFold: single-sequence 속도/scalability 우위. **핵심 성과 2 — 적용 분야**: drug discovery, enzyme engineering, disease research. **핵심 성과 3 — 현재 과제**: protein dynamics, post-translational modifications, protein-ligand interactions 모델링. **핵심 성과 4 — 향후 방향**: protein language models持续发展 + experimental-computational integration. **핵심**: AlphaFold/protein language models comprehensive review — 단백질 AI Scaling Consolidating.

**AlphaFold3 Adversarial Mutations: 35% Confidence Reliability 한계[8]**
J. Feldman et al.(Computational and Structural Biotechnology Journal 2026)의 연구는 AlphaFold3의 adversarial evaluation을 수행한다[8]. **핵심 성과 1 — 40% residues invariance**: 점변이나 결실에도 구조 불변. **핵심 성과 2 — 35% confidence reliability**: confidence metrics가 most accurate structure를 선택하는 비율. **핵심 성과 3 — Template-based pattern matching**: biophysical reasoning보다 memorized templates 의존 가능성. **핵심**: AlphaFold3 adversarial mutations 35% confidence reliability 한계 — 단백질 AI Scaling Consolidating.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Yin review**로 **protein language models 향후 방향**(Synthesizing)을, **Mahtha benchmark**로 **ESMFold dimeric 41% 원인 분석**(Synthesizing)을 제시했다면, 오늘은 **Singh Current Opinion**으로 **protein-ligand data gap으로 인한 실제 약물 발견 적용 한계**(Scaling — 산업 적용 관점)를(new angle), **Yin comprehensive review**를 **Scaling 문맥에서 모델 비교와 적용 분야 정리**로, **Feldman**을 **Scaling 한계 확인**으로 재제시한다. Synthesizing(9/16) → 단백질 AI Scaling(9/17).

실무 함의: 약물 설계팀에서 Singh et al.의 protein-ligand data gap 분석을 참조하여 AlphaFold application strategy를 수립해야 한다.

*출처: [Singh AlphaFold Protein Ligand Drug Discovery Current Opinion Structural Biology 2026](https://consensus.app/papers/details/81a3128ae1615cfeb1bd6544d3ad2f1b/), [Yin AlphaFold Protein Structure Prediction Frontiers Molecular Biosciences 2026](https://consensus.app/papers/details/880e72fce35255c4887811c76849c03a/), [Feldman AlphaFold3 Adversarial Computational Structural Biotechnology 2026](https://consensus.app/papers/details/68f35194db525fb68a7bc8c10546001b/)*

*#ProteinAI #Singh #CurrentOpinion #StructuralBiology #AlphaFold #ProteinLigand #DataGap #HierarchicalFramework #PDB #Frontiers #Yin #FrontiersMolecularBiosciences #AlphaFold3 #RoseTTAFold #ESMFold #ProteinLanguageModels #DeepLearning #Feldman #ComputationalStructuralBiotechnology #Adversarial #35Percent #Confidence #Reliability #40Percent #ResiduesInvariance #TemplateBased #PatternMatching #ScalingDay #DrugDiscovery #ApplicationLimits #DataNeeds #BiomedicalAI #DailyReport*

---

### Post 5: Single-Cell Multi-Omics Foundation Model — FMs vs DMs Performance Gap + Dataset Scaling Plateau Scaling 확인

🧫 **[Single-Cell Multi-Omics] Wang et al.으로 SCMBench 23 methods benchmark — foundation models이 domain-specific models에 미달 + lightweight adaptation strategy 제안(Nature Communications 2026) + DenAdel et al.으로 scFM 22.2M cells × 400 models × 6,400 experiments — dataset scaling plateau 확인, data scaling laws 없음(Nature Methods 2026) + Hou et al.으로 unified scFM framework 13 FMs + 50+ datasets benchmark — classical PCA 여전히 경쟁력(bioRxiv 2026): Single-Cell Foundation Model Scaling(확산) 한계 동시 확인**

Single-Cell Foundation Model의**방법론 한계와 산업 적용 한계(Scaling)**가 출처에서 동시에 확인된다[9][10][11].

**SCMBench: Foundation Models < Domain-Specific Models[9]**
Y. Wang et al.(Nature Communications 2026)의 연구는 SCMBench로 DMs와 FMs를 체계적으로 benchmark한다[9]. **평가 항목**: integration accuracy, biomarker detection, trajectory inference, batch effect correction. **핵심 성과 1 — FMs DMs에 미달**: multi-omics integration에서 foundation models이 최첨단 domain-specific models에 미달. **핵심 성과 2 — Lightweight adaptation strategy**: foundation models 효과를 향상시키는 lightweight adaptation strategy 제안. **핵심 성과 3 — Domain-specific models 우위 확인**: multi-omics integration에서 DMs 우위. **핵심**: SCMBench FMs < DMs + lightweight adaptation — Single-Cell Scaling limitations.

**scFM Dataset Scaling: Plateau + No Data Scaling Laws[10]**
A. DenAdel et al.(Nature Methods 2026)의 연구는 scFM dataset scaling 효과를 6,400개 실험으로 분석한다[10]. **핵심 성과 1 — Scaling plateau**: 현재 방법들이 일부 데이터만으로도 plateau. **핵심 성과 2 — No data scaling laws**: large language models과 달리 scaling laws 없음. **핵심 성과 3 — 균형 강조**: model capacity, dataset size, computational resources 균형 중요. **핵심**: scFM dataset scaling plateau — Single-Cell Scaling methodology limitations.

**Unified scFM Framework: 13 FMs + 50+ Datasets[11]**
S. Hou et al.(bioRxiv 2026)의 연구는 unified computational framework로 13개 FMs를 benchmark한다[11]. **핵심 성과 1 — Classical PCA 여전히 경쟁력**: classical approach remains competitive or even preferable. **핵심 성과 2 — FMs 우위 상황**: pre-trained embeddings가 biologically meaningful structure capture + low-label/transfer-learning scenarios에서 우위. **핵심 성과 3 — Technical barriers 감소**: unified framework로 technical barriers 감소. **핵심**: unified 13 FMs benchmark — classical PCA competitive + FMs advantages in specific scenarios — Single-Cell Scaling Consolidating.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Wang SPARKLE**으로 **RNA leakage correction methodology**(Synthesizing)를, **DenAdel**으로 **dataset scaling plateau**(Synthesizing)를 제시했다면, 오늘은 **Wang SCMBench**를 **Scaling 산업 적용 관점에서 재해석**하여(FMs < DMs의 산업적 함의), **DenAdel plateau**를 **Scaling 한계로**, **Hou**을 **classical PCA + FMs hybrid approach로 재제시**한다. Synthesizing(9/16) → Single-Cell Scaling(9/17).

실무 함의: Single-cell 분석팀에서 Wang et al.의 SCMBench 결과를 참조하여 foundation model vs domain-specific model selection strategy를 수립해야 한다.

*출처: [Wang SCMBench Nature Communications 2026](https://consensus.app/papers/details/24b273d01c1f508c9a20a5a981ac2e3c/), [DenAdel scFM Dataset Scaling Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/), [Hou Unified scFM Framework bioRxiv 2026](https://www.biorxiv.org/content/10.64898/2026.09.01.747412v1)*

*#SingleCell #MultiOmics #FoundationModel #Wang #NatureCommunications #SCMBench #FMs #DMs #LightweightAdaptation #DenAdel #NatureMethods #scFM #DatasetScaling #Plateau #NoScalingLaws #22MillionCells #400Models #6400Experiments #Hou #BioRxiv #UnifiedFramework #13FMs #50Datasets #PCA #Competitive #TransferLearning #ScalingDay #Benchmark #Methodology #Integration #Biomarker #Trajectory #BatchEffect #BiomedicalAI #DailyReport*

---

### Post 6: FDA/규제 — FDA 38 CGTs Orphan 86.8% + Expedited 92.1% + Postmarketing Primary Efficacy 17.9%와 Gene Therapy 산업 확산 현실

🏛️ **[FDA/규제] Shahzad et al.으로 FDA 38개 CGTs 86.8% orphan designation + 92.1% expedited pathway + 73.7% 단일 pivotal study + 44.7% surrogate endpoints solely + postmarketing requirements 73.4% + postmarketing primary efficacy endpoint 17.9%(Clinical Pharmacology Therapeutics 2026) + Oo et al.으로 FDA gene therapy 1998-2025 evolution + platform technology designation + AI/ML mechanistic/model-informed drug development(J Clinical Pharmacology 2026) + Kunanayagam et al.으로 FDA 44 CGTs + 14개 소아 사용 pediatric clinical pharmacology 체계화(Pediatric Research 2026): FDA/규제의 Scaling(확산) 동시 확인**

FDA/규제 영역의**산업 확산과 규제 현실(Scaling)**이 출처에서 동시에 확인된다[12][13][14].

**FDA 38개 CGTs 규제 현황: 73.7% 단일 Pivotal + Postmarketing Primary Efficacy 17.9%[12]**
M. Shahzad et al.(Clinical Pharmacology & Therapeutics 2026)의 연구는 FDA 38개 CGT의 규제 분석을 제시한다[12]. **핵심 성과 1 — 86.8% orphan designation**: 대부분 rare diseases 대상. **핵심 성과 2 — 92.1% expedited pathway**: expedited pathway 활용률 압도적. **핵심 성과 3 — 73.7% 단일 pivotal study**: 단일 연구 기반 승인 다수. **핵심 성과 4 — 44.7% surrogate endpoints solely**: surrogate endpoints만 기반 승인. **핵심 성과 5 — 73.4% postmarketing requirements**: 승인 후 추가 연구 요구 common. **핵심 성과 6 — 17.9% postmarketing primary efficacy endpoint**: 장기 추적에서 1차 임상疗效 종착 분석 17.9%만. **핵심**: FDA CGT 규제 현실 — expedited approval + postmarketing evidence dependence — FDA Scaling Consolidating.

**FDA Gene Therapy Platform Technology Designation: 1998-2025 Evolution[13]**
C. Oo et al.(The Journal of Clinical Pharmacology 2026)의 연구는 FDA gene therapy approvals의演化와 platform technology designation을 분석한다[13]. **승인 역사**: slow start → recent acceleration across RNA-based agents, viral and non-viral in vivo platforms, ex vivo genetically modified cell therapies. **Platform-aligned initiatives**: (i) Plausible mechanism framework, (ii) CMC flexibility initiative, (iii) Advanced manufacturing technologies program. **AI/ML 역할**: dose selection, safety evaluation, durability prediction 지원. **핵심**: FDA gene therapy 1998-2025 evolution + platform technology designation — FDA/규제 Scaling Consolidating.

**CGTs 소아 적용: 44개 승인 + 14개 소아[14]**
S. Kunanayagam et al.(Pediatric Research 2026)의 review는 CGTs의 소아 적용을 분석한다[14]. **승인 현황**: 2025년 5월까지 44개 CGTs 승인( cord blood 제외 14개 소아 사용). **핵심 고려사항**: vector shedding, biodistribution, cellular kinetics — expansion, persistence. **핵심**: FDA CGTs 소아 적용 현황 — FDA/규제 Scaling Consolidating.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Sheikh Kresladi 6개월 follow-up**으로 **real-world evidence 수집**(Synthesizing)을, **Oo/Shahzad FDA**로 **CGT 규제 체계**(Synthesizing)를 제시했다면, 오늘은 **Shahzad Clinical Pharmacology Therapeutics**를 **Scaling 산업 확산 관점에서 재해석**하여(73.7% 단일 pivotal study + 17.9% postmarketing primary efficacy의 산업적 함의), **Oo**를 **Scaling 규제 프레임워크**로, **Kunanayagam**을 **소아 적용 Scaling**으로 재제시한다. Synthesizing(9/16) → FDA Scaling(9/17).

실무 함의: 규제 기획팀에서 Shahzad et al.의 FDA CGT 규제 현실 분석을 참조하여 gene therapy regulatory strategy를 수립해야 한다.

*출처: [Shahzad FDA CGT Assessment Clinical Pharmacology Therapeutics 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Oo FDA Gene Therapy Evolution J Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/), [Kunanayagam FDA CGT Pediatrics Pediatric Research 2026](https://consensus.app/papers/details/1c13548ad5be537a8451eed2c85ecd75/)*

*#FDA #Regulation #GeneTherapy #CGTs #Shahzad #ClinicalPharmacologyTherapeutics #OrphanDesignation #ExpeditedPathway #SingleStudy #SurrogateEndpoints #Postmarketing #17Point9Percent #Oo #JClinicalPharmacology #PlatformTechnology #AI #ML #Mechanistic #ModelInformed #Kunanayagam #PediatricResearch #44CGTs #14Pediatric #ScalingDay #RegulatoryFramework #ExpeditedApproval #PostmarketingEvidence #RareDisease #Pediatric #IndustrialScaling #BiomedicalAI #DailyReport*

---

### Post 7: Spatial Transcriptomics/Cancer Atlas — Pan-Cancer 56 LCPs + Niche_4/11 Prognosis과 Immunotherapy Resistance Prediction Scaling

🗺️ **[Spatial Transcriptomics/Cancer Atlas] Li et al.으로 12개 암종 373 samples · 56개 LCPs · 13개 recurrent niches · Niche_4 macrophage-tumor poor prognosis + Niche_11 better survival + immunotherapy resistance prediction(Cell Reports Medicine 2026) + Zhang et al.으로 SCLC 75 patients · 105 specimens · 600,000+ cells · PIHS-1 pan-immune hotspot survival predictor(Cell Reports Medicine 2026) + Gao et al.으로 gastric cancer Lymphocyte Aggregated Region(LAR) + TLS spatial characterization(Nature Communications 2026): Spatial Transcriptomics/Cancer Atlas Scaling(확산) 동시 확인**

Spatial Transcriptomics/Cancer Atlas의**면역치료 저항성 예측과 대규모 적용(Scaling)**이 출처에서 동시에 확인된다[15][16][17].

**Pan-Cancer Spatial Microenvironment: 56 LCPs · 13 Niches · Niche_4/11[15]**
J. Li et al.(Cell Reports Medicine 2026)의 연구는 12개 암종의 spatial transcriptomics를 수행한다[15]. **연구 규모**: 373 samples · 12 cancer types. **핵심 성과 1 — 56개 LCPs(local cellular programs)**: tumor spatial microenvironment 고유한 발현 프로그램. **핵심 성과 2 — 13개 recurrent niches**: niche-shared + niche-specific ligand-receptor interactions. **핵심 성과 3 — Niche_4 poor prognosis**: macrophages + tumor cells colocalization — poor prognosis + immunotherapy resistance. **핵심 성과 4 — Niche_11 better survival**: macrophages colocalized with immune cells — better survival + treatment response. **핵심 성과 5 — Immunotherapy resistance prediction**: Niche_4/11 기반 resistance prediction model 가능성. **핵심**: Pan-cancer 56 LCPs + Niche_4/11 prognosis + immunotherapy resistance prediction — Spatial Scaling Consolidating.

**SCLC Spatial Map: 600,000+ Cells · PIHS-1 Survival Predictor[16]**
Z. Zhang et al.(Cell Reports Medicine 2026)의 연구는 SCLC의 lymph node metastasis에서 spatial transcriptomics를 수행한다[16]. **연구 규모**: 75 SCLC patients · 105 specimens · 600,000+ cells · CosMx Spatial Molecular Imager. **핵심 성과 1 — 3개 LNM-enriched malignant subclusters**: distinct metabolic + angiogenic programs + immune exclusion features. **핵심 성과 2 — PIHS-1 (pan-immune hotspot 1)**: 독립적 survival predictor. **핵심 성과 3 — Vascular-immune crosstalk**: endothelial cells가 cytotoxic T cells와 functional perivascular niches 형성. **핵심**: SCLC spatial map + PIHS-1 survival predictor — Spatial Scaling Consolidating.

**Gastric Cancer LAR: Lymphocyte Aggregated Region + TLS[17]**
S. Gao et al.(Nature Communications 2026)의 연구는 gastric cancer의 spatial microenvironment를 특성화한다[17]. **핵심 성과 1 — Lymphocyte Aggregated Region(LAR)**: lymphocyte aggregates + tertiary lymphoid structures(TLS). **핵심 성과 2 — T Cell Activation Association**: Within LAR에서 naive T cell abundance와 T cell activation-associated pathways 연관. **핵심 성과 3 — PD1+CD27+CD8 + CD70+LAMP3+ DC proximity**: Group A 인접 종양 영역에서 클러스터. **핵심**: Gastric cancer LAR + TLS spatial characterization — Spatial Scaling Consolidating.

**9월 16일 Synthesizing vs 오늘의 차별점**: 9/16이 **Li Niche_4/11**으로 **immunotherapy resistance prediction outlook**(Synthesizing)을, **Gao LAR**으로 **Group A/B immune activation 차이**(Synthesizing)를 제시했다면, 오늘은 **Li 56 LCPs + Niche_4/11**을 **Scaling 대규모 적용 관점에서 재해석**하여(12개 암종 373 샘플의 대규모 데이터로 확산), **Zhang 600K cells**을 **Scaling 대규모 코호트 atlas**로, **Gao LAR**을 **Scaling 임상 적용 확산**으로 재제시한다. Synthesizing(9/16) → Spatial Scaling(9/17).

실무 함의: 암면역 치료 연구팀에서 Li et al.의 Niche_4/11 framework를 참조하여 immunotherapy resistance prediction strategy를 수립해야 한다.

*출처: [Li Pan-cancer Spatial Transcriptomics Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/), [Zhang SCLC Spatial Transcriptomics Cell Reports Medicine 2026](https://consensus.app/papers/details/dc8b02ddbfcb574fb92b18cf8aed1117/), [Gao Gastric Cancer LAR Nature Communications 2026](https://consensus.app/papers/details/46f3d71acd2e5b6e8816d6886aa485d9/)*

*#SpatialTranscriptomics #CancerAtlas #Li #CellReportsMedicine #PanCancer #LCP #Niche #Niche4 #Niche11 #PoorPrognosis #BetterSurvival #ImmunotherapyResistance #Zhang #SCLC #CellReportsMedicine #600000Cells #CosMx #LNM #PIHS1 #SurvivalPredictor #VascularImmune #Crosstalk #Gao #GastricCancer #LAR #TLS #TertiaryLymphoidStructures #PD1 #CD27 #CD8 #DendriticCells #ScalingDay #TumorMicroenvironment #LargeScale #Atlas #ClinicalTranslation #BiomedicalAI #DailyReport*

---

### Post 8: 오늘의 요약 — Scaling(확산) Day 핵심 정리

🧬 **Scaling(확산) Day — 9월 17일 핵심 정리**

오늘 Scaling Day의 핵심 성과 3가지:

**1. PE First-in-Human + PE7/PE8 Evolution: CRISPR 임상 이행 大規模 확산**
Lushington et al.(Molecular Therapy 2026)으로 prime editing first-in-human functional restoration + promising safety profile[1]. Yu et al.(Advanced Science 2026)으로 PE1→PE7 evolution + architecture innovations[2]. Mengistu et al.(The Journal of Gene Medicine 2026)으로 CRISPR-Cas systems Phase I-III trial outcomes 종합[3]. CRISPR Scaling — **실제 인체 실험 milestone + 기술 architecture 확산 + 임상 시험 전체 스펙트럼 산업화**.

**2. DrugCLIP 10M× Virtual Screening + GenomeScreenDB: AI Drug Discovery 산업 인프라 확산**
Jia et al.(Science 2026)으로 DrugCLIP contrastive learning genome-wide virtual screening — docking 대비 10,000,000× 속도 + GenomeScreenDB 10,000 proteins × 500M compounds 사전 계산 DB 공개[4]. Singh et al.(Current Opinion in Structural Biology 2026)으로 AlphaFold-like models protein-ligand data gap — 약물 발견 적용 위한 PDB data 확대 필요성[5]. AI drug discovery Scaling — **초고속 가상 선별 infrastructure + 대규모 사전 계산 DB 확산 + 데이터 필요성 한계 인식**.

**3. FDA 38 CGTs 규제 현실 + Pan-Cancer Spatial Atlas: 규제·임상 데이터 산업 확산**
Shahzad et al.(Clinical Pharmacology Therapeutics 2026)으로 FDA 38 CGTs 86.8% orphan + 92.1% expedited + 73.7% 단일 pivotal study + postmarketing primary efficacy 17.9%[12]. Li et al.(Cell Reports Medicine 2026)으로 12개 암종 56 LCPs + Niche_4/11 + immunotherapy resistance prediction[15]. FDA/Spatial Scaling — **규제 산업 확산 현실 확인 + 대규모 공간적 종양 미세환경 atlas 확산**.

**핵심 인사이트:** Synthesizing(9/16)에서 확인된 통합 프레임워크들이 오늘 **실제 임상 이행 milestone(PE first-in-human), 대규모 산업 인프라(DrugCLIP 10M× + GenomeScreenDB), 규제 표준화(FDA 38 CGTs 규제 현실), 대규모 atlas 확산(Pan-cancer 56 LCPs)**으로 Scaling하고 있다.

*출처: [Lushington PE First-in-Human Molecular Therapy 2026](https://consensus.app/papers/details/79a173589b1d58598fd94fa334768881/), [Yu PE Evolution Advanced Science 2026](https://consensus.app/papers/details/475fc3cbbe805f5e915626f08d2da039/), [Mengistu CRISPR Cas Systems Journal of Gene Medicine 2026](https://consensus.app/papers/details/6f6486cb3f3f5baab6f2ebaf3e67c39d/), [Jia DrugCLIP Science 2026](https://consensus.app/papers/details/5acea8f3221b5acfa37184d1fe50da28/), [Singh AlphaFold Current Opinion Structural Biology 2026](https://consensus.app/papers/details/81a3128ae1615cfeb1bd6544d3ad2f1b/), [Shahzad FDA CGT Clinical Pharmacology Therapeutics 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Li Pan-cancer Spatial Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/), [Zhang SCLC Spatial Cell Reports Medicine 2026](https://consensus.app/papers/details/dc8b02ddbfcb574fb92b18cf8aed1117/), [Gao Gastric Cancer Nature Communications 2026](https://consensus.app/papers/details/46f3d71acd2e5b6e8816d6886aa485d9/), [Yin AlphaFold Frontiers Molecular Biosciences 2026](https://consensus.app/papers/details/880e72fce35255c4887811c76849c03a/), [Feldman AlphaFold3 Adversarial Computational 2026](https://consensus.app/papers/details/68f35194db525fb68a7bc8c10546001b/), [Wang SCMBench Nature Communications 2026](https://consensus.app/papers/details/24b273d01c1f508c9a20a5a981ac2e3c/), [DenAdel scFM Dataset Scaling Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/), [Hou Unified scFM Framework bioRxiv 2026](https://www.biorxiv.org/content/10.64898/2026.09.01.747412v1), [Mao AI Drug Discovery Frontiers Pharmacology 2026](https://consensus.app/papers/details/ca5321af0bb656e095b6108766231764/), [Oo FDA Gene Therapy J Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/), [Kunanayagam FDA CGT Pediatrics Pediatric Research 2026](https://consensus.app/papers/details/1c13548ad5be537a8451eed2c85ecd75/)*

*#ScalingDay #Scaling #Summary #Lushington #MolecularTherapy #PrimeEditing #FirstInHuman #Yu #AdvancedScience #PE7 #PE8 #Mengistu #JournalOfGeneMedicine #CRISPRCas #Jia #DrugCLIP #Science #10MillionTimes #Faster #GenomeScreenDB #Singh #CurrentOpinion #AlphaFold #ProteinLigand #DataGap #Shahzad #FDA #CGTs #Orphan #Expedited #SingleStudy #Postmarketing #17Point9Percent #Li #PanCancer #LCP #Niche4 #Niche11 #Zhang #SCLC #PIHS1 #Gao #LAR #TLS #Wang #SCMBench #DenAdel #Hou #Mao #PhaseI #PhaseII #TargetValidity #Oo #PlatformTechnology #Kunanayagam #Pediatric #Synthesizing #IndustrialScaling #ClinicalTranslation #BiomedicalAI #DailyReport*

---

## Sources (1차 출처, 10개 이상 URL)

[1] [Lushington et al. — A primer on prime: A prime editing update from advances to first-in-human trial. Molecular Therapy 2026](https://consensus.app/papers/details/79a173589b1d58598fd94fa334768881/)

[2] [Yu et al. — Evolution of Prime Editing: Enhancing Efficiency and Expanding Capacity. Advanced Science 2026](https://consensus.app/papers/details/475fc3cbbe805f5e915626f08d2da039/)

[3] [Mengistu et al. — CRISPR–Cas Systems in Human Disease Therapy: Advances, Clinical Applications, Limitations, and Future Directions. The Journal of Gene Medicine 2026](https://consensus.app/papers/details/6f6486cb3f3f5baab6f2ebaf3e67c39d/)

[4] [Jia et al. — Deep contrastive learning enables genome-wide virtual screening. Science 2026](https://consensus.app/papers/details/5acea8f3221b5acfa37184d1fe50da28/)

[5] [Singh et al. — More Protein-Ligand data is needed for AlphaFold-like Models to enable drug discovery. Current Opinion in Structural Biology 2026](https://consensus.app/papers/details/81a3128ae1615cfeb1bd6544d3ad2f1b/)

[6] [Mao et al. — Artificial intelligence in drug discovery: from algorithmic foundations to clinical translation. Frontiers in Pharmacology 2026](https://consensus.app/papers/details/ca5321af0bb656e095b6108766231764/)

[7] [Yin et al. — Protein structure prediction powered by artificial intelligence: from biochemical foundations to practical applications. Frontiers in Molecular Biosciences 2026](https://consensus.app/papers/details/880e72fce35255c4887811c76849c03a/)

[8] [Feldman et al. — Adversarial Sequence Mutations in AlphaFold and ESMFold Reveal Nonphysical Structural Invariance. Computational and Structural Biotechnology Journal 2026](https://consensus.app/papers/details/68f35194db525fb68a7bc8c10546001b/)

[9] [Wang et al. — SCMBench: benchmarking domain-specific and foundation models for single-cell multi-omics data integration. Nature Communications 2026](https://consensus.app/papers/details/24b273d01c1f508c9a20a5a981ac2e3c/)

[10] [DenAdel et al. — Evaluating the role of pre-training dataset size and diversity on single-cell foundation model performance. Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/)

[11] [Hou et al. — A unified computational framework for benchmarking single-cell foundation models. bioRxiv 2026](https://www.biorxiv.org/content/10.64898/2026.09.01.747412v1)

[12] [Shahzad et al. — Assessment of United States Food and Drug Administration Approval of Cell and Gene Therapies. Clinical Pharmacology and Therapeutics 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/)

[13] [Oo et al. — FDA Gene Therapy Approvals (1998–2025): Current Status, Regulatory Evolution, and Future Directions. The Journal of Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/)

[14] [Kunanayagam et al. — Clinical pharmacology insights from recent cell and gene therapy approvals relevant to pediatrics. Pediatric Research 2026](https://consensus.app/papers/details/1c13548ad5be537a8451eed2c85ecd75/)

[15] [Li et al. — Pan-cancer analysis of spatial transcriptomics reveals heterogeneous tumor spatial microenvironment. Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/)

[16] [Zhang et al. — Single-cell spatial transcriptomics reveals tumor microenvironment heterogeneity in primary and lymph node-metastatic small cell lung cancer. Cell Reports Medicine 2026](https://consensus.app/papers/details/dc8b02ddbfcb574fb92b18cf8aed1117/)

[17] [Gao et al. — A spatially resolved atlas of gastric cancer characterises a lymphocyte-aggregated region. Nature Communications 2026](https://consensus.app/papers/details/46f3d71acd2e5b6e8816d6886aa485d9/)