Life Sciences / Regulatory Review 🧬

📌 Navigate

📋 Exec Summary

Q2 2024 saw AlphaFold 3 redefine computational biology by expanding from protein-only to multi-molecular complex prediction, while FDA AI/ML device clearances continued accelerating through the quarter. Later full-year 2024 analysis put the total at 168 devices, with radiology still dominant. The GLP-1 market hit $13.84B as Lilly's tirzepatide gained share, EU MDR/IVDR notified body bottlenecks worsened, and the Recursion-Exscientia merger ($688M) was telegraphed as AI drug discovery consolidation began in earnest.

📊 What Moved

📈 Trend Arcs

Arc 1: AI-Enabled Device Regulatory Maturation

Velocity: Accelerating

The FDA moved from ad hoc review of AI/ML devices toward a systematic framework. The PCCP (Predetermined Change Control Plan) concept — allowing manufacturers to pre-specify how an AI model will be updated post-clearance — advanced through 2024, with the final AI-enabled-device guidance still ahead in 2025. During Q2, approximately 16.7% of AI/ML device summaries already referenced PCCPs. This is the regulatory mechanism that makes continuously-learning medical AI commercially viable.

Where it stands at quarter close: The FDA has signaled that iterative AI improvement and regulatory oversight are not mutually exclusive. Builders with clean PCCP narratives gain a structural advantage in review timelines.

Arc 2: Computational Biology Leaps from Structure to Interaction

Velocity: Accelerating

AlphaFold 2 solved single-protein structure prediction. AlphaFold 3 solved interaction prediction — proteins with ligands, DNA, RNA, and ions. This is the difference between knowing what a lock looks like and knowing which keys fit. For drug discovery, the implication is that structure-based virtual screening can now operate on complexes rather than isolated targets, compressing hit-finding timelines.

Where it stands at quarter close: AF3's diffusion-based architecture is published but the full model weights are restricted to Isomorphic Labs' commercial pipeline. Academic access is via the AlphaFold Server. Builders in AI-driven drug discovery must decide whether to build on the AF3 paradigm or develop competing approaches.

Arc 3: GLP-1 Supply Chain as Market Bottleneck

Velocity: Steady

Demand for GLP-1 drugs exceeded manufacturing capacity throughout Q2. Lilly and Novo Nordisk both committed multi-billion-dollar manufacturing expansions, but new capacity takes 3-5 years to come online. Compounding pharmacies attempted to fill the gap, creating regulatory tension. The bottleneck was not scientific, regulatory, or demand-side — it was purely industrial.

Where it stands at quarter close: The GLP-1 market is supply-constrained, not demand-constrained. This inverts the normal biopharma dynamic and creates an unusual window for competitors who can manufacture at scale.

🗺️ Landscape Shift

🧭 Regulatory Direction of Travel

FDA — U.S.

The FDA's Center for Devices and Radiological Health (CDRH) continued establishing itself as the most AI-literate regulatory body globally. Key Q2 signals:

• PCCP advancement: The predetermined change control plan framework was still maturing in Q2, providing a mechanism for AI/ML devices to update post-clearance without a new submission for each iteration. This is the single most important regulatory enabler for continuously-learning medical AI.
• AI/ML clearance velocity: Later full-year analysis put 168 total clearances in 2024, with 94.6% via 510(k). The review pipeline is functioning, not bottlenecked.
• Transparency expectations: Later full-year analysis showed 29.2% of AI/ML device summaries reporting both sensitivity and specificity and 15.5% providing demographic performance data. The bar is rising but remains inconsistent.

EU — MDR/IVDR

The European regulatory environment continued to deteriorate for medtech innovation:

• Notified body capacity: The structural constraint is not resolving. Extended transition periods have bought time but not expanded throughput.
• EUDAMED delays: The central database required for full compliance remained in phased rollout, creating uncertainty for manufacturers planning submissions.
• Disproportionate burden on SMEs: The European Commission's own targeted evaluation acknowledged that compliance costs are disproportionate, suggesting future simplification — but simplification timelines are measured in years, not quarters.
• Regulation 2024/1860: New supply chain notification requirements and revised IVDR deadlines added complexity during the quarter.

Net direction: The U.S. is building regulatory infrastructure that enables AI medical devices. The EU is struggling to maintain baseline device market access. For builders, U.S.-first regulatory strategy has never been more clearly advantaged.

💰 Funding & Deal Pattern

🔍 Counter-Narrative

• The consensus: AlphaFold 3 is a transformational leap for structural biology. The reality: Unlike AF2, full model weights are not openly available — Isomorphic Labs retains commercial rights. If the most powerful tool in structural biology is gated behind a commercial entity, the open-science ecosystem that made AF2 transformational may fragment rather than advance.
• The consensus: Later full-year 2024 analysis showed 168 FDA AI/ML clearances and proves regulatory maturation. The reality: Only 29.2% reported both sensitivity and specificity; just 15.5% included demographic performance data. High clearance rate with inconsistent transparency risks market saturation of poorly characterized AI devices, which could trigger regulatory backlash across the entire category.

📐 Builder's Benchmark

👀 What to Watch

📎 Sources