Gene delivery with AAV at scale: capsid Innovation, emerging alternatives, and the road to commercial reality
08 Sept 2026
Next Gen Vector Engineering & Delivery
- From empirical to predictive capsid engineering — Directed evolution has driven most capsid gains so far, but AI/ML-guided and structure-based design are maturing fast. Are we ready to predict tropism and liver de-targeting rather than screen for it — and what would it take to trust a computationally designed capsid in the clinic?
- The immunity ceiling — Pre-existing neutralizing antibodies exclude many patients, and treatment-induced immunity blocks re-dosing. Which approach actually breaks this: immune-evasive capsids, transient immunosuppression, capsid switching — or does it push whole indications toward non-viral delivery?
- Is manufacturing the real rate-limiter? — Full-to-empty ratios, scalability, and cost of goods remain stubborn hurdles. Are our biggest constraints now in vector biology, or downstream in CMC and characterization — and how much clinical potential is stranded by manufacturing rather than design?
- When do alternatives displace AAV? — LNPs and other non-viral platforms increasingly enable extrahepatic delivery, easier redosing, and simpler manufacturing. Where do they already have the edge, and where does AAV retain a durable advantage?
- HEK293 vs. Sf9/baculovirus — which platform scales? — Transient triple transfection in HEK293 offers human-relevant post-translational profiles but strains at scale, while Sf9/baculovirus scales more readily with trade-offs in capsid quality and comparability. As programs move toward commercial volumes, is one platform emerging as the default — or does the choice stay indication- and product-specific?
- Has analytics kept pace with the clinic? — AAV characterization is uniquely complex — full/empty/partial ratios, aggregation, residuals, potency, and genome integrity — and standardized, orthogonal methods still lag behind the pace of clinical development. Where is the analytical gap most acute, and is CMC characterization now the hidden bottleneck to approvals and comparability?
