Drug Discovery12

Artificial Intelligence • Gene Therapy • Radioligand Theranostics

🤖 Artificial Intelligence

FDA and EMA Expand AI-Enabled Medical Device Oversight and Knowledge Mining Tools

FDA updated its AI-Enabled Medical Devices list on March 4, documenting over 1,000 approved devices with new transparency requirements for foundation models. Simultaneously, EMA extended its "Scientific Explorer" tool to include EPARs for initial marketing authorization, operationalizing AI for regulatory assessors. Explicit warnings were issued regarding liability for AI-enabled scams and deepfakes.

Regulatory fda.gov — AI-Enabled Medical Devices

🧬 Gene Therapy

FDA and ARM Announce April 2026 Workshop on Pediatric Cell and Gene Therapy Trials

CBER and the Alliance for Regenerative Medicine will host a workshop on April 9, 2026, to address challenges in pediatric CGT trials. Topics include ethical thresholds for "prospect of direct benefit," timing for presymptomatic enrollment, and biomarker-driven designs. This signals FDA's intent to formalize pathways for early intervention in rare diseases.

Regulatory / Workshop fda.gov — Advancing Pediatric CGT Trials

⚛️ Radioligand Theranostics

No Major Regulatory Updates; Real-World Data Supports Earlier-Line Use

No new approvals or major trial readouts this week. The field continues to digest Feb 2026 data showing Pluvicto's median PFS of 15.8 months when used after one ARPI vs. 12.7 months after multiple ARPIs. This reinforces the clinical rationale for earlier RLT sequencing rather than salvage use.

Clinical Context (Contextual Update)

📡 Expert Signals

EMA AI Observatory 2024 Annual Report: Governance is Key

EMA's first AI Observatory report highlights accelerating LLM adoption for document processing but emphasizes non-negotiable human oversight. It offers a blueprint for responsible AI: aggressive internal use wrapped in strict validation and governance.

Regulatory Report ema.europa.eu — AI Observatory

Nature Microbiology: "Navigating the Promise and Pitfalls of AI"

Editorial warns that while ML aids antimicrobial discovery, LLMs still produce vague outputs requiring manual curation. The thesis: "machines in the loop, not humans in the loop"—AI should augment, not replace, scientific hypothesis generation.

Journal / Editorial nature.com — Promise and Pitfalls of AI

Nature: Releasing Open-Weight AI in Steps to Alleviate Risks

Policy piece argues for staged releases of open-weight models to balance scientific access with safety. It calls for peer review and provenance tracking, reminding regulated sectors that "open-source" does not equal "unregulated."

Policy Perspective nature.com — Open-Weight AI Risks

📰 In Case You Missed It

1. FDA Draft Guidance on AI-Enabled Device Lifecycle Management (Jan 2026)

Proposes lifecycle frameworks for adaptive AI, emphasizing predetermined change control plans. Essential reading for developers of AI diagnostics and clinical decision support tools.

Regulatory fda.gov — AI Device Lifecycle

2. FDA Outlines Flexible CMC Expectations for CGTs (Jan 2026)

Announced risk-based CMC approaches, including phased data expectations and flexibility on PPQ lots, reducing early-stage manufacturing burdens for ultra-rare therapies.

Regulatory fda.gov — Flexible CMC Requirements

3. MHRA Approves Earlier-Line Pluvicto for PSMA+ mCRPC (Feb 2026)

Authorized for use after ARPI progression but before taxane chemotherapy, based on PSMAfore data showing 59% risk reduction in radiographic progression/death vs. ARPI switch.

Regulatory novartis.com — MHRA Approval

4. U.S. Treasury Issues Financial Services AI Risk Management Framework

Aligns with NIST AI RMF but tailored for finance. Serves as a de facto governance reference for any regulated industry deploying high-stakes AI models.

Policy Framework treasury.gov — AI Risk Framework

5. Nature: AI Tools Design Genomes and Synthetic Viruses

Reports on "generative biology" moving from analysis to design, creating novel proteins and synthetic viruses. Highlights dual-use risks and the need for robust biosafety threat modeling.

Science / Ethics nature.com — AI Designing Genomes

6. Nature: First "AI Societies" Taking Shape

Researchers are training AI agents to mimic human social interactions. While early, this suggests future utility in simulating clinical trial recruitment dynamics and payer decision-making.

Emerging Tech nature.com — AI Societies

7. Nature: FDA Clears First Cellular Rejuvenation Trial (Feb 2026)

Life Biosciences' ER-100 trial targets partial epigenetic reprogramming. Signals regulatory openness to "rejuvenation" concepts if safety and control logic are rigorously proven.

Clinical Trial nature.com — Cellular Rejuvenation

🎯 Practical Implications

  • Device Regulation: Review the Jan 2026 FDA draft guidance immediately; stricter documentation for algorithm validation and change control is crystallizing.
  • Pediatric Trials: Monitor the April 9 FDA-ARM workshop for new "prospect of direct benefit" interpretations that will shape future IND/BLA reviews.
  • Internal AI Use: Adopt the EMA AI Observatory model—treat AI as infrastructure requiring critical human cross-checks and documented validation.
  • Transparency: Proactively disclose foundation model use in 510(k) summaries to differentiate from opaque developers.
  • Research Validation: Implement staged validation for open-weight models in research; verify outputs before scientific reliance.