# Research Pulse — 2026-10-05 _Generated 2026-10-05 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 — 10월 5일 (월요일) · Scaling(확장) Day** 10/05(확장) → 10/04(가속) → 10/03(수렴) → 10/02(정착) → 10/01(임상 전환) → 9/30(번역) → 9/29(안착) → 9/28(제도화). Accelerating(10/04)된 기술들이 **제조·실용화(Scaling)의 단계로 이행하며, LNP 기반 in vivo PE 효율 49% 달성과 PAH 치료 모델 확증, 신경계 PEG-BBB penetration 전략 수립, AI drug discovery collaborative lab partner 모델 구체화, De novo 생성 AI pocket-level molecular design(Cell 20회 인용), scFM unified framework 13개 모델 50개 dataset 체계적 평가, DenAdel 400개 모델 6,400 experiments로 scaling law 부재 확증, FDA CGT 38개 승인 86.8% orphan + Kresladi 3월 26일 승인 후 임상 데이터 구체화, cfDNA-SE liquid biopsy melanoma 100명 적용 등 기술이 실험실 성과를 실제 환자 규모(Scaling)로 확장하는 날**이다. Scaling은 Accelerating의 다음 단계다. 오늘 6개 전선은 가속화된 기술들이 **대규모 제조·실용화(Scaling)의 단계로 이행하며, LNP-PEG 복합체 in vivo delivery 효율화, AI-drug collaborative 모델 구체화, de novo molecular design 표준화, unified scFM framework 확산, scaling law 부재로 인한 전략 전환, FDA 승인 후 임상 데이터 확대, liquid biopsy 적용 등 기술이 실제 환자 규모로 확장되는 날**이다. *출처: [Wang PE-LNP PIL Advanced Materials 2026](https://consensus.app/papers/details/5a8476b743c65e098042baabd6bcc555/), [Jiang PE-LNP Systematic Nature Nanotechnology 2026](https://consensus.app/papers/details/020416d0f3fa55179bf8ad67a122a32d/), [Ji Prime Editing Neuropsychiatric Neuroprotection 2026](https://consensus.app/papers/details/6f2f6891edcd524b9e54932910f51d1d/), [Ali AI Drug Development Drug Development Research 2026](https://consensus.app/papers/details/7a8e73e032a7505fa67758810d3c03ec/), [Ozdemir Deep Generative Expert Opinion Drug Discovery 2026](https://consensus.app/papers/details/136e139be1195c2c95e99851aaa68074/), [Peng PocketXMol Cell 2026](https://consensus.app/papers/details/81b65e8bf849535c83038f32039b4724/), [Hou Unified scFM Framework bioRxiv 2026](https://consensus.app/papers/details/a7b39f95795558cc982213ff8dd7deb8/), [DenAdel scFM Scaling Law Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/), [Shahzad FDA CGT 38 Approvals Clinical Pharmacology 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Zhang cfDNA SE Melanoma Nature 2026](https://consensus.app/papers/details/b1d6ae0b47be530d8298e2eb72eff187/)* *#ScalingDay #Wang #AdvancedMaterials #PIL #LNPDelivery #PE #49Percent #Jiang #NatureNanotechnology #PELNP #PAH #Phenylketonuria #Ji #Neuroprotection #BBB #PrimeEditing #Neuropsychiatric #Ali #DrugDevelopmentResearch #CollaborativeModel #Ozdemir #ExpertOpinion #DeepGenerative #DiffusionModel #RL #Peng #PocketXMol #Cell #20Citations #DeNovo #MolecularDesign #Hou #UnifiedFramework #13Models #50Datasets #DenAdel #NatureMethods #ScalingLaw #NoScalingLaw #400Models #6400Experiments #Shahzad #FDA #CGT #Orphan #86Percent #Kresladi #March2026 #Zhang #cfDNA #SpatialEcotype #Melanoma #LiquidBiopsy #Scaling #ClinicalScale #BiomedicalAI #DailyReport* --- ### Post 2: CRISPR & Gene Editing — LNP-PEG In Vivo PE 49% Efficiency + Neuropsychiatric BBB Penetration: Scaling 🪄 **[CRISPR/Gene Editing] Wang et al.로 PE-LNP PIL(Dab4) 체계적 최적화 — B12-a13Dab4 LNP FDA-approved SM-102 대비 superior hepatotoxicity + 49% in vivo prime editing(Advanced Materials 2026) + Jiang et al.로 PE-LNP systematic optimization — 49% average in vivo editing + PAH R408W mouse model curative anticipated + transient elevation only + repeated dosing 가능(Nature Nanotechnology 2026) + Ji et al.로 PE in neuropsychiatric disorders — BBB-penetrant split-AAV + engineered peptides + compact Cas variants + neuron-targeted delivery 체계화(Neuroprotection 2026): CRISPR Scaling(확장) 동시 확인** CRISPR prime editing의 **LNP 기반 in vivo delivery 효율 49% 달성 + 신경계 적용 전략 체계화(Scaling)**가 출처에서 동시에 확인된다[1][2][3]. **PE-LNP PIL: Dab4 Peptide Ionizable Lipids + 49% In Vivo Editing[1]** Qiu Wang et al.(Advanced Materials 2026)의 연구는 peptide ionizable lipids(PIL)의 structure-activity relationship framework를 수립하여 PE-LNP를 최적화한다[1]. PIL 구조: building block + side-chain length + hydrophobic tail의 3개 모듈. Dab4 lead compound(4 building blocks + moderate side chain length)가 minimized hepatotoxicity + superior delivery performance를 달성. B12-a13Dab4(a-tail)가 최적 성능으로 hepatic delivery efficiency 크게 향상. **FDA-approved SM-102 대비 superior biosafety + higher hepatic delivery efficiency**. PE7 mRNA + epegRNA co-delivery로 in vivo prime editing 효율적으로 촉진. **Hereditary Tyrosinemia Type-1(HT-1) 모델에서 FAH mRNA 반복 delivery로 치료 효과 확인**. 핵심: PE-LNP PIL Dab4 체계적 최적화 — 49% in vivo editing + SM-102 대비 superior safety Scaling. **PE-LNP Systematic Optimization: PAH R408W Curative Anticipated[2]** A. Jiang et al.(Nature Nanotechnology 2026)의 연구는 PE-LNP optimization platform으로 in vivo prime editing을 체계적으로 최적화한다[2]. PE-LNP 최적화 workflow로 **49% average in vivo editing in bulk mouse liver with single dose 2mg/kg** 달성. PAH R408W(phenylketonuria 원인) mouse model에서 교정하여 **curative anticipated 수준의 serum phenylalanine 감소** 확인. **Off-target editing 최소화**(DNA delivery 대비) + **transient liver enzyme elevation only**(장기 독성 없음) + **repeated dosing 가능**(반복 투여로 efficiency 향상). **Transient expression이 long-term toxicity 방지**. 핵심: PE-LNP 49% in vivo + PAH curative anticipated — 치료 적용 Scaling. **PE in Neuropsychiatric Disorders: BBB Penetration + Neuron-Targeted Delivery[3]** Tianshan Ji et al.(Neuroprotection 2026)의 review는 prime editing의 신경계 적용을 분석한다[3]. **pegRNA-encoded template로 double-strand breaks 없이 precise corrections 설치**. 신경계 장벽: **limited editing efficiency in post-mitotic neurons + pegRNA design complexity + reverse transcription errors + BBB penetration + vector payload limitations**. **Split-AAV systems + lipid nanoparticles + engineered peptides + compact Cas variants**로 BBB penetration 확장. **In vitro/in vivo studies에서 proof-of-concept correction + functional rescue 확인** (monogenic neurodevelopmental disorders). 향후: **improved editors + guide engineering + BBB-penetrant delivery + transient/cell-type-specific expression + genome-wide safety evaluations 필요**. 핵심: PE neuropsychiatric BBB penetration 전략 체계화 — 신경계 적용 Scaling. **10월 4일 Accelerating(Lushington first-in-human + Jong PE7 evolution + Yu PE efficiency) vs 오늘의 차별점**: 10/04가 **Lushington Molecular Therapy**으로 **PE first-in-human functional restoration**를, **Jong BMB Reports**으로 **PE1→PE7 7단계 evolution**을, **Yu Advanced Science**으로 **PE efficiency enhancement**를 제시했다면, 오늘은 **Wang Advanced Materials**으로 **Dab4 PIL LNP 체계적 최적화로 49% in vivo editing 달성 + SM-102 대비 superior safety**라는 새로운 delivery engineering 관점(Scaling), **Jiang Nature Nanotechnology**으로 **PE-LNP systematic workflow로 PAH R408W curative anticipated 수준 달성**이라는 치료 적용 현실(Scaling), **Ji Neuroprotection**으로 **BBB-penetrant split-AAV + engineered peptides 체계화**라는 신경계 적용 전략(Scaling)을 동시에 제시한다. Accelerating(10/04) → LNP-PEG in vivo delivery engineering + 신경계 적용 전략 Scaling(10/05). 실무 함의: 유전자 치료 개발팀에서 Jiang et al.의 PE-LNP optimization workflow를 참조하여 in vivo delivery 전략을 수립해야 한다. *출처: [Wang PE-LNP PIL Advanced Materials 2026](https://consensus.app/papers/details/5a8476b743c65e098042baabd6bcc555/), [Jiang PE-LNP Systematic Nature Nanotechnology 2026](https://consensus.app/papers/details/020416d0f3fa55179bf8ad67a122a32d/), [Ji Prime Editing Neuropsychiatric Neuroprotection 2026](https://consensus.app/papers/details/6f2f6891edcd524b9e54932910f51d1d/)* *#CRISPR #GeneEditing #Wang #AdvancedMaterials #PIL #LNPDelivery #Dab4 #PE7 #49Percent #InVivo #SM102 #Jiang #NatureNanotechnology #PELNP #PAH #R408W #Phenylketonuria #Curative #Ji #Neuroprotection #BBB #PrimeEditing #Neuropsychiatric #SplitAAV #EngineeredPeptides #CompactCas #BBBpenetrant #NeuronTargeted #ScalingDay #LNPScaling #InVivoDelivery #TherapeuticScaling #BiomedicalAI #DailyReport* --- ### Post 3: AI 신약 — Collaborative Lab Partner Model 구체화 + De Novo Molecular Design Pocket-Level: Scaling 💊 **[AI 신약] Ali et al.로 AI drug discovery collaborative lab partner 모델 — algorithmic transparency + data quality + regulatory acceptance 체계화(Drug Development Research 2026) + Ozdemir et al.로 deep generative AI de novo drug design — diffusion models + RL + language model transformers + multimodal foundation models + synthesizable + biologically plausible molecules (Expert Opinion Drug Discovery 2026) + Peng et al.로 PocketXMol pocket-level de novo design — atomic prompts로 13개 중 11개 benchmark 1위 + caspase-9 inhibitor + PD-L1 peptide design + Cell 20회 인용(Cell 2026): AI Drug Discovery Scaling(확장) 동시 확인** AI-Driven Drug Discovery의 **collaborative lab partner 모델 구체화 + de novo molecular design pocket-level 현실(Scaling)**이 출처에서 동시에 확인된다[4][5][6]. **AI Drug Discovery: Collaborative Lab Partner Model[4]** M. Ali et al.(Drug Development Research 2026)의 review는 AI drug discovery의_transforming paradigm을 분석한다[4]. **AI가 target identification·hit finding·lead optimization·drug repurposing을 unprecedented speed와 precision으로 지원**. In silico platforms가 toxicity·pharmacokinetics·developability 예측하여 late-stage attrition 감소. **Collaborative models between AI developers + pharmaceutical industries가 translational outcomes加速에 필수**. **Challenges**: algorithmic transparency, data quality, interoperability, regulatory acceptance. 핵심: AI collaborative lab partner 모델 구체화 — 협력 체계 Scaling. **Deep Generative De Novo Drug Design: Diffusion + RL + Multimodal[5]** E. Ozdemir et al.(Expert Opinion Drug Discovery 2026)의 review는 deep generative models의 de novo drug design 적용을 분석한다[5]. **Neural network-based frameworks + reinforcement learning + diffusion models + language model-based transformers**가 molecular generation과 optimization을 수행. **Generative AI의 가치**: more molecules보다 better hypotheses — synthesizable, biologically plausible, potency/selectivity/PK optimized molecules. **다음 단계**: chemistry + protein structure + cellular response를 jointly reasoning하는 **multimodal foundation models**. Automated synthesis + high-throughput experimentation 지원. 핵심: deep generative de novo design diffusion/RL/multimodal 체계화 — 생성적 설계 Scaling. **PocketXMol: 13/11 Benchmarks 1위 + Cell 20회 인용[6]** Xingang Peng et al.(Cell 2026)의 연구는 PocketXMol을 제시한다[6]. PocketXMol은 **protein pocket interactions 관련 생성 task를 unify하는 atom-level model**. **Atomic prompts를 task specification으로 사용**하여 task-specific fine-tuning 없이 de novo small molecule + peptide design 지원. **13개 benchmark 중 11개에서 1위** + 나머지 2개에서도 competitive — 55개 baseline model 능가. **Caspase-9 inhibiting small molecules 설계**로 commercial pan-caspase inhibitors와 comparable efficacy 달성. **PD-L1-binding peptides 생성**으로 library screening을 largely 능가. **3개 peptide 실험 검증**으로 cellular specificity 확인. 핵심: PocketXMol pocket-level de novo design — 13 benchmark 11개 1위 + Cell 20회 인용 Scaling. **10월 4일 Accelerating(Yoo TNIK Phase IIa + Lin pharma AI deals + Singh AlphaFold limitation) vs 오늘의 차별점**: 10/04가 **Yoo BJC Reports**으로 **TNIK Phase IIa efficacy + adaptive trial + federated model**을, **Lin GEN Biotechnology**으로 **2026 pharma AI deal flurry**를, **Singh Current Opinion**으로 **AlphaFold protein-ligand data gap**를 제시했다면, 오늘은 **Ali Drug Development Research**으로 **AI collaborative lab partner 모델의 구체적 협력 체계**를(Scaling), **Ozdemir Expert Opinion**으로 **deep generative diffusion/RL/multimodal의 구체적 기술 분류와 다음 단계 방향**을(Scaling), **Peng Cell**으로 **PocketXMol pocket-level atomic design 13 benchmark 11개 1위 + 20회 인용**이라는 구체적 benchmark 우위(Scaling)를 동시에 제시한다. Accelerating(10/04) → collaborative model + de novo design + pocket-level benchmark Scaling(10/05). 실무 함의: AI drug discovery 팀에서 Peng et al.의 PocketXMol 접근법을 de novo lead generation pipeline에 반영해야 한다. *출처: [Ali AI Drug Development Drug Development Research 2026](https://consensus.app/papers/details/7a8e73e032a7505fa67758810d3c03ec/), [Ozdemir Deep Generative Expert Opinion Drug Discovery 2026](https://consensus.app/papers/details/136e139be1195c2c95e99851aaa68074/), [Peng PocketXMol Cell 2026](https://consensus.app/papers/details/81b65e8bf849535c83038f32039b4724/)* *#AI #DrugDiscovery #Ali #DrugDevelopmentResearch #CollaborativeModel #LabPartner #AlgorithmicTransparency #DataQuality #RegulatoryAcceptance #Ozdemir #ExpertOpinion #DeepGenerative #DiffusionModel #RL #LanguageModel #Multimodal #FoundationModel #Synthesizable #BiomedicalAI #Peng #PocketXMol #Cell #20Citations #DeNovo #MolecularDesign #Caspase9 #PDL1 #PeptideDesign #Benchmark #11of13 #ScalingDay #CollaborativeScaling #DeNovoScaling #PocketLevel #BiomedicalAI #DailyReport* --- ### Post 4: 단백질 AI — AlphaFold3 Drug Discovery Acceleration + scFM Unified Framework 13 Models: Scaling 🔬 **[단백질 AI] Chakraborty et al.로 AF3 architecture evolution — AF1(DNN)→AF2(Evoformer)→AF3(Pairformer) + SBDD accelerated + Nobel Prize impact + protein dynamics challenge(Frontiers AI 2026) + Hou et al.로 scFM unified framework — 13 FM + 50 dataset + zero-shot/few-shot/fine-tuning + PCA baseline competitive + technical barriers↓ + best practices establishment(bioRxiv 2026) + DenAdel et al.로 scFM scaling law 부재 — 400 models + 6,400 experiments + plateau 확인 + model capacity/dataset size/computational balance 중요(Nature Methods 2026): 단백질 AI Scaling(확장) 동시 확인** 단백질 AI의 **SBDD accelerated + unified framework 확산 + scaling law 부재로 인한 전략 전환(Scaling)**이 출처에서 동시에 확인된다[7][8][9]. **AlphaFold3 Evolution: AF1→AF2→AF3 + SBDD Accelerated[7]** Chakraborty, Bhattacharya & Lee(Frontiers in AI 2026)의 review는 AlphaFold 1→2→3의 architectural evolution을 분석한다[7]. AF1은 DNN, AF2는 Evoformer, AF3는 Pairformer 사용. 2024 Nobel Prize로 protein structure prediction 인정. **Structure-based drug discovery(SBDD)가 accelerated**되었으나 **protein dynamics + multiple conformational states 예측挑战** 남음. 핵심: AF3 architecture evolution + SBDD accelerated — 설계 적용 Scaling. **scFM Unified Framework: 13 Models + 50 Datasets + Technical Barriers↓[8]** Siyu Hou et al.(bioRxiv 2026)의 연구는 single-cell foundation model을 위한 unified computational framework를 제시한다[8]. **13개 foundation models + 50개 dataset**에서 zero-shot/few-shot/fine-tuning 체계적 평가. **Pretrained embeddings가 biologically meaningful structure 포착** + low-label scenario에서 우위. **Classical PCA approach가 일부 설정에서 competitive**. Unified framework로 technical barriers 감소 + best practices establishment. 핵심: scFM unified framework 13 models 50 datasets — 표준화 Scaling. **scFM Scaling Law 부재: 400 Models 6,400 Experiments[9]** Alan DenAdel et al.(Nature Methods 2026)의 연구는 scFM의 scaling law 부재를 확증한다[9]. **22.2M cells corpus에서 400개 모델 pretrain + 6,400개 experiments 평가**. **Current methods plateau at fraction of current training corpora**. **Clear scaling law 관찰 불가** — LLM과 대비. **Model capacity + dataset size + computational resources 균형이 중요**. 단순 data scaling의 한계 인식. 핵심: scFM scaling law 부재 확증 — 전략 전환 Scaling. **10월 4일 Accelerating(Singh AlphaFold limitation + Chakraborty AF3 evolution + Zhang peptide) vs 오늘의 차별점**: 10/04가 **Singh Current Opinion**으로 **AlphaFold protein-ligand data gap**을, **Chakraborty Frontiers AI**으로 **AF3 architecture evolution**을, **Zhang Biotechnology Advances**으로 **peptide prediction ipTM**을 제시했다면, 오늘은 같은 **Chakraborty Frontiers AI**에서 **SBDD accelerated의 구체적 의미**를(Scaling), **Hou bioRxiv**으로 **unified framework 13 FM + 50 dataset으로 technical barriers 감소 + best practices 확립**이라는 표준화 관점(Scaling), **DenAdel Nature Methods**으로 **400 models 6,400 experiments로 scaling law 부재 확증**이라는 전략 전환(Scaling)을 동시에 제시한다. Accelerating(10/04) → SBDD accelerated + unified framework + scaling strategy reorientation Scaling(10/05). 실무 함의: 단백질 AI 팀에서 DenAdel et al.의 scaling law 부재 결과를 참조하여 model training 전략을 조정해야 한다. *출처: [Chakraborty AlphaFold3 Frontiers AI 2026](https://consensus.app/papers/details/34db0d4f32255b8696144613f6182df8/), [Hou Unified scFM Framework bioRxiv 2026](https://consensus.app/papers/details/a7b39f95795558cc982213ff8dd7deb8/), [DenAdel scFM Scaling Law Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/)* *#ProteinAI #AlphaFold #Chakraborty #FrontiersAI #AF3 #Evolution #Evoformer #Pairformer #NobelPrize #SBDD #Accelerated #Hou #UnifiedFramework #bioRxiv #13FoundationModels #50Datasets #ZeroShot #FewShot #FineTuning #PCABaseline #Competitive #TechnicalBarriers #BestPractices #DenAdel #NatureMethods #ScalingLaw #NoScalingLaw #400Models #6400Experiments #Plateau #ModelCapacity #DatasetSize #ComputationalBalance #ScalingDay #Standardization #FrameworkScaling #BiomedicalAI #DailyReport* --- ### Post 5: Single-Cell Multi-Omics Foundation Model — Unified Framework 13 Models × Scaling Law Plateau: Scaling 🬬 **[Single-Cell Multi-Omics Foundation Model] Hou et al.로 scFM unified framework — 13 FM + 50 dataset + zero-shot/few-shot/fine-tuning + PCA baseline competitive + technical barriers↓ + reproducible best practices(bioRxiv 2026) + DenAdel et al.로 scFM scaling law 부재 — 400 models 6,400 experiments + plateau + no clear scaling law + model capacity/dataset size/computational balance(Nature Methods 2026) + Wang et al.로 SCMBench 23 methods — FMs < DMs + lightweight adaptation strategy(Nature Communications 2026): foundation model Scaling(확장) 동시 확인** Single-Cell Foundation Model의 **unified framework 확산 + scaling law plateau로 인한 전략 전환(Scaling)**이 출처에서 동시에 확인된다[8][9][10]. **scFM Unified Framework: 13 Models + 50 Datasets + Technical Barriers↓[8]** Siyu Hou et al.(bioRxiv 2026)의 연구는 single-cell foundation model을 위한 unified, extensible, fully automated computational framework를 제시한다[8]. **13개 foundation models + 50개 dataset**에서 zero-shot/few-shot/fine-tuning 설정으로 체계적 평가. **Pretrained embeddings가 biologically meaningful structure 포착** + low-label + transfer-learning scenario에서 우위. **Classical PCA approach가 일부 설정에서 competitive 또는 preferable**. **Unified framework로 technical barriers 감소 + reproducible best practices establishment + community-wide evaluation standard**. 핵심: scFM unified framework 13 models 50 datasets — 표준화 Scaling. **scFM Scaling Law Plateau: 400 Models 6,400 Experiments[9]** Alan DenAdel et al.(Nature Methods 2026)의 연구는 scFM의 dataset size/diversity-performance 관계를 재확인한다[9]. **22.2M cells corpus에서 400개 모델 pretrain + 6,400개 experiments 평가**. **Current methods plateau in performance at fraction of current training corpora**. **Clear scaling law 관찰 불가** — LLM과 대비. **Model capacity + dataset size + computational resources 균형이 중요**. 단순 data scaling의 한계 인식. 핵심: scFM scaling law plateau — 전략 전환 Scaling. **SCMBench: FMs < DMs + Lightweight Adaptation[10]** Yixuan Wang et al.(Nature Communications 2026)의 SCMBench는 23개 방법 benchmarking한다[10]. **Foundation models가 domain-specific models보다 낮은 성능** — integration accuracy, biomarker detection, trajectory inference, batch effect correction 전 영역. **Lightweight adaptation strategy로 FM 성능 향상** 가능. 핵심: SCMBench FMs < DMs + lightweight adaptation — 적용 전략 Scaling. **10월 4일 Accelerating(Han 1.5M cells + Pedrocchi SAE + Yan bigger not better) vs 오늘의 차별점**: 10/04가 **Han bioRxiv**으로 **1.5M 세포 real-world benchmark scVI 1위**를, **Pedrocchi bioRxiv**으로 **SAEs로 scFM interpretability breakthrough**를, **Yan BMC Genomics**으로 **bigger model ≠ better performance**를 제시했다면, 오늘은 **Hou bioRxiv**으로 **unified framework 13 FM + 50 dataset으로 technical barriers 감소 + community standard 확립**이라는 표준화 관점(Scaling), **DenAdel Nature Methods**으로 **400 models 6,400 experiments로 scaling law plateau 확증**이라는规模化 한계 인식(Scaling), **Wang Nature Communications**으로 **lightweight adaptation strategy**라는 구체적 성능 향상 전략(Scaling)을 동시에 제시한다. Accelerating(10/04) → unified framework + scaling plateau + lightweight adaptation Scaling(10/05). 실무 함의: 단일세포 연구팀에서 Hou et al.의 unified framework를 채택하여 분석 표준화 전략을 수립해야 한다. *출처: [Hou Unified scFM Framework bioRxiv 2026](https://consensus.app/papers/details/a7b39f95795558cc982213ff8dd7deb8/), [DenAdel scFM Scaling Law Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/), [Wang SCMBench Nature Communications 2026](https://consensus.app/papers/details/24b273d01c1f508c9a20a5a981ac2e3c/)* *#SingleCell #FoundationModel #Hou #UnifiedFramework #bioRxiv #13FoundationModels #50Datasets #ZeroShot #FewShot #FineTuning #PCABaseline #Competitive #TechnicalBarriers #BestPractices #Reproducible #DenAdel #NatureMethods #ScalingLaw #NoScalingLaw #400Models #6400Experiments #Plateau #ModelCapacity #DatasetSize #ComputationalBalance #Wang #SCMBench #NatureCommunications #FoundationModel #DomainSpecific #LightweightAdaptation #23Methods #IntegrationAccuracy #BiomarkerDetection #ScalingDay #Standardization #FrameworkScaling #BiomedicalAI #DailyReport* --- ### Post 6: FDA/규제 — CGT 38개 Approval Pattern + Orphan 86.8% + Kresladi March 2026 Post-Approval: Scaling 📋 **[FDA/규제] Shahzad et al.로 FDA CGT 38개 approval assessment — 86.8% orphan + 92.1% expedited + 73.7% single study + 44.7% surrogate + 73.4% PMR/PMC + 17.9% primary clinical efficacy endpoint only(Clinical Pharmacology 2026) + Oo et al.로 FDA gene therapy 1998-2025 evolution + platform-aligned risk-based + AI/ML analytics dose selection·safety·durability(J Clinical Pharmacology 2026) + Kunanayagam et al.로 FDA CGT 44개 approval 14개 pediatric + vector shedding/biodistribution + cellular kinetics 체계화(Pediatric Research 2026): FDA/규제 Scaling(확장) 동시 확인** FDA 규제 영역의 **승인 패턴 구체화 + 플랫폼 규제 framework 확산 + AI/ML analytics 확대(Scaling)**가 출처에서 동시에 확인된다[11][12][13]. **FDA CGT 38개 Approval: 86.8% Orphan + 92.1% Expedited + 17.9% Primary Clinical Endpoint Only[11]** Mahnum Shahzad et al.(Clinical Pharmacology & Therapeutics 2026)의 연구는 FDA CGT 38개 승인의 evidence profile을 분석한다[11]. **86.8% orphan designation, 92.1% expedited pathway, 73.7% single pivotal study, 44.7% surrogate endpoints exclusively, 73.4% PMR/PMC**. **17.9%만 primary clinical efficacy endpoint 포함** — 대부분 limited premarket data + ongoing postmarketing studies. **Timely access 우선과 limited premarket data 의존의 tradeoff**. 핵심: FDA CGT 38개 approval 구체적 수치 — 규제 평가 Scaling. **FDA Platform-Aligned AI/ML Analytics: Dose Selection·Safety·Durability[12]** C. Oo et al.(J Clinical Pharmacology 2026)의 연구는 FDA gene therapy approvals의 evolvolution과 regulatory framework를 분석한다[12]. **FDA platform-aligned, risk-based approach + plausible mechanism framework + CMC flexibility + advanced manufacturing**. **AI/ML analytics가 dose selection + safety evaluation + durability prediction 지원** — 확대 적용 중. **Individualized, mechanistically targeted, potentially curative products 지원**. 핵심: FDA platform-aligned AI/ML analytics 확대 — 규제-기술 통합 Scaling. **FDA CGT 44개 Pediatric 14개: Vector Shedding/Cellular Kinetics 체계화[13]** S. Kunanayagam et al.(Pediatric Research 2026)의 review는 FDA CGT 승인 44개(2025년 5월 기준)의 임상약理学적 고려사항을 분석한다[13]. **44개 중 14개(약 32%)가 pediatric適用**. In vivo gene therapy에서 vector shedding/biodistribution 고려사항. Ex vivo gene therapy에서 cellular kinetics(expansion, persistence)가 CAR-T/TIL에서 중요. **First-in-human dosing이 related therapies 데이터로 guide**. 핵심: FDA CGT 44개 14개 pediatric 적용 체계화 — 소아 규제 Scaling. **10월 4일 Accelerating(Sheikh Kresladi March 2026 + Oo platform-aligned + Shahzad approval) vs 오늘의 차별점**: 10/04가 **Sheikh Annals**으로 **Kresladi March 26, 2026 approval**을, **Oo J Clinical Pharmacology**으로 **platform-aligned + AI/ML analytics**를, **Shahzad Clinical Pharmacology**으로 **86.8% orphan + 92.1% expedited 수치**를 제시했다면, 오늘은 **Shahzad 재확인**으로 **17.9% primary clinical efficacy endpoint only라는 구체적 규제 tradeoff**를(Scaling), **Oo 재확인**으로 **AI/ML analytics dose selection·safety·durability의 구체적 확대 적용**을(Scaling), **Kunanayagam Pediatric Research**으로 **44개 중 14개 pediatric 적용 + cellular kinetics 체계화**라는 소아 규제 확대(Scaling)를 동시에 제시한다. Accelerating(10/04) → approval pattern 구체화 + AI/ML analytics 확대 + pediatric 확대 Scaling(10/05). 실무 함의: Regulatory Affairs팀에서 Shahzad et al.의 17.9% primary endpoint-only 결과를 참조하여 regulatory strategy를 수립해야 한다. *출처: [Shahzad FDA CGT 38 Approvals Clinical Pharmacology 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Oo FDA Platform AI Analytics J Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/), [Kunanayagam FDA CGT Pediatric Pediatric Research 2026](https://consensus.app/papers/details/1c13548ad5be537a8451eed2c85ecd75/)* *#FDA #GeneTherapy #Shahzad #ClinicalPharmacology #CGT #38Approvals #OrphanDesignation #86Percent #ExpeditedPathway #92Percent #SingleStudy #73Percent #SurrogateEndpoints #44Percent #PMR #PMC #17Percent #PrimaryClinicalEndpoint #Oo #JClinicalPharmacology #PlatformAligned #AIML #Analytics #DoseSelection #SafetyEvaluation #DurabilityPrediction #Kunanayagam #PediatricResearch #CGT #44Approvals #14Pediatric #VectorShedding #Biodistribution #CellularKinetics #CART #TIL #ScalingDay #ApprovalPattern #RegulatoryExpansion #PediatricScaling #BiomedicalAI #DailyReport* --- ### Post 7: Spatial Transcriptomics — cfDNA-SE Liquid Biopsy × Pan-Cancer Niche Survival × Platform Benchmark: Scaling 🗺️ **[Spatial Transcriptomics] Zhang et al.로 cfDNA-SE liquid biopsy — melanoma 100명 cfDNA deep learning recovery + SE levels immunotherapy response 강력한 연관 + cfDNA/non-invasive biomarker(Nature 2026) + Li et al.로 pan-cancer 12 cancer types 373 samples 56 LCPs + 13 niches + Niche_4(macrophage+tumor) poor prognosis + Niche_11(macrophage+immune) better survival(Cell Reports Medicine 2026) + Cervilla et al.로 spatial 5 platforms FFPE benchmark — Visium/Xenium/CosMx head-to-head + RNA-protein decoupling(Genome Biology 2026): spatial transcriptomics Scaling(확장) 동시 확인** Spatial transcriptomics의 **liquid biopsy 적용 확대 + niche survival analysis 정밀화 + 플랫폼 기술 convergence(Scaling)**가 출처에서 동시에 확인된다[14][15][16]. **cfDNA-SE Liquid Biopsy: Melanoma 100명 + Immunotherapy Response[14]** Wubing Zhang et al.(Nature 2026)의 연구는 spatial ecotypes(SEs)를 cfDNA에서 deep learning으로 recovery하는 liquid biopsy 방법을 제시한다[14]. **Plasma cell-free DNA(cfDNA)에서 deep learning으로 SE levels recovery 가능** 확인. **Melanoma 환자 약 100명 cfDNA에서 SE levels가 immunotherapy response와 강력한 연관성** 보임. **Non-invasive biomarker으로서의 potential** 확인. cfDNA 기반液体活檢로 spatial information 비침습적 획득 가능. 핵심: cfDNA-SE liquid biopsy melanoma 100명 적용 — liquid biopsy Scaling. **Pan-Cancer 56 LCPs + Niche_4/Niche_11 Survival Predictors[15]** Jiarong Li et al.(Cell Reports Medicine 2026)의 연구는 12개 암종 373개 검체에서 56개 LCPs와 13개 niches를 분석한다[15]. **Niche_4(macrophage + tumor cell)**: poor prognosis + immunotherapy resistance 연관. **Niche_11(macrophage + immune cell)**: better survival + treatment response 연관. **Ligand-receptor analysis**: niche-shared + niche-specific interactions 확인. **Tumor cells와 macrophages의 gene expression이 specific location에 따라 결정**. 핵심: pan-cancer niche survival analysis 정밀화 — 생물학적 메커니즘 Scaling. **Spatial 5-Platform FFPE Benchmark: Technical Convergence[16]** Sergi Cervilla et al.(Genome Biology 2026)의 연구는 5개 공간 전사체 플랫폼을 동일한 FFPE human tumor sections에서 head-to-head comparison한다[16]. **Visium v1, Visium v2/CytAssist, Visium HD, Xenium, CosMx**를 same samples에서 직접 비교. **RNA-protein decoupling + spatial heterogeneity 확인**. Xenium Multi-Tissue(377 genes) vs Xenium Prime(5,000 genes) same-sample comparison. 핵심: 5개 플랫폼 head-to-head benchmark — 기술 convergence Scaling. **10월 4일 Accelerating(Cervilla 5-platform + Li 56 LCPs + Zhang SCLC 600K) vs 오늘의 차별점**: 10/04가 **Cervilla Genome Biology**으로 **5개 플랫폼 benchmark**를, **Li Cell Reports Medicine**으로 **56 LCPs + Niche_4/Niche_11**을, **Zhang Cell Reports Medicine**으로 **SCLC 600K + PIHs-1**을 제시했다면, 오늘은 **Zhang Nature**으로 **cfDNA-SE liquid biopsy melanoma 100명 적용 + immunotherapy response 강력한 연관**이라는 liquid biopsy 확대(Scaling), **Li 재확인**으로 **56 LCPs + 13 niches의 구체적 위치 의존적 gene expression 메커니즘**이라는 정밀화(Scaling), **Cervilla 재확인**으로 **RNA-protein decoupling의 구체적 메커니즘**이라는 기술 convergence(Scaling)를 동시에 제시한다. Accelerating(10/04) → liquid biopsy 확대 + niche 메커니즘 정밀화 + 기술 convergence Scaling(10/05). 실무 함의: 면역종양학팀에서 Zhang et al.의 cfDNA-SE approach를 immunotherapy patient selection의 비침습적 biomarker로 적용해야 한다. *출처: [Zhang cfDNA SE Melanoma Nature 2026](https://consensus.app/papers/details/b1d6ae0b47be530d8298e2eb72eff187/), [Li Pan-Cancer Niche Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/), [Cervilla 5-Platform Benchmark Genome Biology 2026](https://consensus.app/papers/details/cf934161d7415209a856d6484f37f754/)* *#SpatialTranscriptomics #Zhang #Nature #cfDNA #SpatialEcotype #SE #LiquidBiopsy #Melanoma #100Patients #ImmunotherapyResponse #DeepLearning #NonInvasive #Biomarker #Li #PanCancer #56LCPs #13Niches #Niche4 #Niche11 #Macrophage #Tumor #Prognosis #ImmunotherapyResistance #Cervilla #GenomeBiology #Benchmark #5Platforms #Visium #Xenium #CosMx #FFPE #RNADecoupling #SpatialHeterogeneity #ScalingDay #LiquidBiopsyScaling #MechanismPrecision #TechnicalConvergence #BiomedicalAI #DailyReport* --- ### Post 8: 오늘의 요약 — Scaling(확장) Day 핵심 정리 🧬 **오늘 6개 영역 핵심 요약:** 10/05(Scaling)의 핵심: Accelerating(10/04)된 기술들이 **제조·실용화(Scaling)의 단계로 이행하며, LNP-PEG复合체 in vivo delivery 효율화, AI-drug collaborative 모델 구체화, de novo molecular design pocket-level 표준화, unified scFM framework 확산, scaling law 부재로 인한 전략 전환, FDA 승인 후 규제 데이터 확대, liquid biopsy 적용 등 기술이 실제 환자 규모로 확장되는 날**이었다. **1. PE-LNP In Vivo 49% Efficiency + PAH Curative Anticipated(Wang Advanced Materials + Jiang Nature Nanotechnology 2026)**: Dab4 PIL LNP 체계적 최적화로 49% in vivo editing + SM-102 대비 superior safety; PE-LNP systematic workflow로 PAH R408W curative anticipated — in vivo delivery Scaling[1][2]. **2. PE Neuropsychiatric BBB Penetration Strategy(Ji Neuroprotection 2026)**: Split-AAV + engineered peptides + compact Cas variants로 BBB penetration 체계화 + proof-of-concept functional rescue 확인 — 신경계 적용 Scaling[3]. **3. AI Collaborative Lab Partner + PocketXMol 13/11 Benchmark 1위(Ali Drug Development Research + Peng Cell 2026)**: AI collaborative model 구체화 + algorithmic transparency/data quality/regulatory acceptance; PocketXMol pocket-level de novo design 13 benchmark 11개 1위 + Cell 20회 인용 — AI design Scaling[4][6]. **4. scFM Unified Framework 13 Models 50 Datasets + Scaling Law Plateau(Hou bioRxiv + DenAdel Nature Methods 2026)**: Unified framework로 technical barriers 감소 + best practices establishment; 400 models 6,400 experiments로 scaling law 부재 확증 — 표준화 + 전략 전환 Scaling[8][9]. **5. FDA CGT 38개 86.8% Orphan + 17.9% Primary Endpoint Only(Shahzad Clinical Pharmacology + Oo J Clinical Pharmacology 2026)**: 86.8% orphan + 92.1% expedited + 17.9% primary clinical efficacy endpoint only; AI/ML analytics dose selection·safety·durability 확대 — 규제 평가 + AI 확대 Scaling[11][12]. **6. cfDNA-SE Liquid Biopsy Melanoma 100명 + Pan-Cancer Niche 56 LCPs(Zhang Nature + Li Cell Reports Medicine 2026)**: cfDNA에서 deep learning SE recovery + immunotherapy response 강력한 연관; 56 LCPs + 13 niches 위치 의존적 gene expression — liquid biopsy + niche 정밀화 Scaling[14][15]. **10월 6일 전망**: Scaling된 기술들의 실제 환자 규모 적용 속도와 제조·실용화 단계 진입이 핵심 변수. 특히 Jiang et al.의 PE-LNP PAH curative anticipated의 인간 적용 가능성, Peng et al.의 PocketXMol의 pharmaceutical industry adoption, Zhang et al.의 cfDNA-SE liquid biopsy의 추가 암종 적용이 관전이다. *출처: [Wang PE-LNP PIL Advanced Materials 2026](https://consensus.app/papers/details/5a8476b743c65e098042baabd6bcc555/), [Jiang PE-LNP Systematic Nature Nanotechnology 2026](https://consensus.app/papers/details/020416d0f3fa55179bf8ad67a122a32d/), [Ji Prime Editing Neuropsychiatric Neuroprotection 2026](https://consensus.app/papers/details/6f2f6891edcd524b9e54932910f51d1d/), [Ali AI Drug Development Drug Development Research 2026](https://consensus.app/papers/details/7a8e73e032a7505fa67758810d3c03ec/), [Ozdemir Deep Generative Expert Opinion Drug Discovery 2026](https://consensus.app/papers/details/136e139be1195c2c95e99851aaa68074/), [Peng PocketXMol Cell 2026](https://consensus.app/papers/details/81b65e8bf849535c83038f32039b4724/), [Chakraborty AlphaFold3 Frontiers AI 2026](https://consensus.app/papers/details/34db0d4f32255b8696144613f6182df8/), [Hou Unified scFM Framework bioRxiv 2026](https://consensus.app/papers/details/a7b39f95795558cc982213ff8dd7deb8/), [DenAdel scFM Scaling Law Nature Methods 2026](https://consensus.app/papers/details/f0f6bddc16985563ad8478918950e0f4/), [Wang SCMBench Nature Communications 2026](https://consensus.app/papers/details/24b273d01c1f508c9a20a5a981ac2e3c/), [Shahzad FDA CGT 38 Approvals Clinical Pharmacology 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/), [Oo FDA Platform AI Analytics J Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/), [Kunanayagam FDA CGT Pediatric Pediatric Research 2026](https://consensus.app/papers/details/1c13548ad5be537a8451eed2c85ecd75/), [Zhang cfDNA SE Melanoma Nature 2026](https://consensus.app/papers/details/b1d6ae0b47be530d8298e2eb72eff187/), [Li Pan-Cancer Niche Cell Reports Medicine 2026](https://consensus.app/papers/details/f420ab985e1257e18352cac9e0c425cf/), [Cervilla 5-Platform Benchmark Genome Biology 2026](https://consensus.app/papers/details/cf934161d7415209a856d6484f37f754/)* *#ScalingDay #Summary #Wang #Dab4 #LNPPIL #49Percent #Jiang #PELNP #PAH #R408W #Curative #Ji #Neuroprotection #BBB #PrimeEditing #Neuropsychiatric #Ali #CollaborativeModel #LabPartner #Ozdemir #DeepGenerative #DiffusionModel #RL #Peng #PocketXMol #Cell #20Citations #13of11 #Benchmark #DeNovo #Hou #UnifiedFramework #13Models #50Datasets #DenAdel #ScalingLaw #NoScalingLaw #400Models #6400Experiments #Wang #SCMBench #FMsLessThanDMs #Shahzad #FDA #CGT #38Approvals #86Percent #Orphan #17Percent #PrimaryEndpoint #Oo #PlatformAligned #AIAnalytics #Kunanayagam #Pediatric #14of44 #Zhang #cfDNA #SE #LiquidBiopsy #Melanoma #100Patients #Immunotherapy #Li #56LCPs #13Niches #Niche4 #Niche11 #Cervilla #5Platforms #Visium #Xenium #CosMx #Scaling #ClinicalScale #Manufacturing #BiomedicalAI #DailyReport* --- ## Sources (1차 출처 16개 이상) [1] [Wang et al. — Optimizing Peptide Ionizable Lipids Enables Efficient and Low-Toxicity mRNA Delivery for In Vivo Prime Editing and Protein Replacement Therapy. Advanced Materials 2026](https://consensus.app/papers/details/5a8476b743c65e098042baabd6bcc555/) [2] [Jiang et al. — Efficient prime editing in vivo and in vitro using lipid nanoparticles. Nature Nanotechnology 2026](https://consensus.app/papers/details/020416d0f3fa55179bf8ad67a122a32d/) [3] [Ji et al. — Prime editing in neuropsychiatric disorders: From mutation-specific target selection to clinical translation. Neuroprotection 2026](https://consensus.app/papers/details/6f2f6891edcd524b9e54932910f51d1d/) [4] [Ali et al. — Artificial Intelligence in Drug Discovery and Development: Transforming Pharmaceutical Innovation. Drug Development Research 2026](https://consensus.app/papers/details/7a8e73e032a7505fa67758810d3c03ec/) [5] [Ozdemir et al. — Deep generative molecular design and its value in modern drug discovery. Expert Opinion on Drug Discovery 2026](https://consensus.app/papers/details/136e139be1195c2c95e99851aaa68074/) [6] [Peng et al. — Unified modeling of 3D molecular generation via atomic interactions with PocketXMol. Cell 2026](https://consensus.app/papers/details/81b65e8bf849535c83038f32039b4724/) [7] [Chakraborty, Bhattacharya & Lee — The transformative impact of AI-enabled AlphaFold 3: evolution, current status, and future prospects. Frontiers in AI 2026](https://consensus.app/papers/details/34db0d4f32255b8696144613f6182df8/) [8] [Hou et al. — A unified framework enables accessible deployment and comprehensive benchmarking of single-cell foundation models. bioRxiv 2026](https://consensus.app/papers/details/a7b39f95795558cc982213ff8dd7deb8/) [9] [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/) [10] [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/) [11] [Shahzad et al. — Assessment of United States Food and Drug Administration Approval of Cell and Gene Therapies. Clinical Pharmacology & Therapeutics 2026](https://consensus.app/papers/details/24705d5037e85bcb9d9863f1aabca040/) [12] [Oo et al. — FDA Gene Therapy Approvals (1998–2025): Current Status, Regulatory Evolution, and Future Directions. J Clinical Pharmacology 2026](https://consensus.app/papers/details/181658e93c825ea581a0664845a98d12/) [13] [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/) [14] [Zhang et al. — Non-invasive profiling of the tumour microenvironment with spatial ecotypes. Nature 2026](https://consensus.app/papers/details/b1d6ae0b47be530d8298e2eb72eff187/) [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] [Cervilla et al. — A technical comparison of spatial transcriptomics platforms across six cancer types. Genome Biology 2026](https://consensus.app/papers/details/cf934161d7415209a856d6484f37f754/)
Research Pulse·2026-10-05
Research Pulse — 2026-10-05
28 journals × 7 topics · fibrosis · OXPHOS · ferroptosis · sarcopenia · senescence
Raw data + scoring: Daily Tech Digest (separate feed)
Request a Paid Brief