Lena Dolgikh

• Leveraging foundation models to navigate complex protein systems (multi-domain, membrane-embedded, inducedproximity targets) that exceed the limits of single-structure or single-model approaches
• Combining AI-driven protein design with experimental structure generation to resolve ambiguity in binding modes, conformational states, and interaction interfaces
• Closing the loop between AI predictions and wet-lab validation by continuously retraining models on construct performance, structural success, and functional readouts

Following a series of perspectives from structural and computational experts, this interactive session will shift the focus from presentation to participation. Through structured, topic-led roundtable rotations, attendees will explore where predictive modelling delivers real value, where it falls short, and how experimental data quality influences model performance.

Participants will compare workflows, discuss bottlenecks in data generation and validation, and share practical insights on integrating modelling with structural and biophysical evidence to improve confidence in discovery decisions.