AI Accelerates AAV Capsid Engineering

Traditional AAV capsid engineering is hampered by limitations in library size and time-consuming experimental iterations, often taking over three years to develop a new variant. AI-driven platforms, however, can reduce this timeline to about one year by using predictive models to design and screen vast virtual libraries for properties like tissue specificity and immune evasion before synthesis. Companies like Latus Bio are leveraging large in-vivo datasets from non-human primate studies to train these predictive models, covering over 100 million unique "delivery zip codes" (route-capsid-tissue-cell type). The transition to AI-driven design necessitates a robust digital infrastructure, a significant challenge in gene therapy manufacturing where data is often fragmented across disparate systems like LIMS, MES, and ERPs. This lack of standardized data architecture hinders the development of effective machine learning models. Initiatives are now focused on creating unified data ecosystems to support continued process verification (CPV) and enable technologies like digital twins for bioprocess optimization. Digital twins—virtual replicas of physical bioreactors—are becoming central to optimizing AAV production by simulating process modifications before implementation. This allows for virtual testing of parameters like agitation speed or aeration rates, accelerating process scale-up and reducing the risk of costly batch failures. Automation platforms, such as Orbit, are being developed to directly link these digital models with lab equipment, creating a fully automated workflow from model calibration to optimized process execution. Beyond the capsid, AI is being used to optimize the entire AAV production system, including the design of tissue-specific promoters and tools to silence gene of interest (GOI) expression during manufacturing to reduce toxicity. A key manufacturing hurdle is the removal of empty viral capsids, which can impact product safety and efficacy. While anion exchange chromatography is a common method, AI can help optimize upstream processes to reduce the empty capsid rate from the start. The biotech funding environment has shifted significantly since the peak in 2021, with investors now more cautious and selective. While the cell and gene therapy sector saw a funding drop of 83% from 2021 to 2024, significant investments are still being made in companies with validated science and clear advantages, particularly in AAV technology. This climate places greater emphasis on CDMOs that offer integrated services, from development to manufacturing, to de-risk investments for smaller biotechs. For technical leaders aiming for executive roles, the transition requires a fundamental shift from problem-solving to people management and strategic thinking. Success depends less on deep technical expertise and more on the ability to develop emotional intelligence, manage cross-functional teams, and align technical decisions with broader business objectives. This often involves unlearning a purely technical focus to embrace strategic leadership and organizational dynamics.