Advancing regulatory variant effect prediction with AlphaGenome - Natasha Latysheva, Senior Research Engineer, Google DeepMind

98% of the human genome does not encode proteins, and these vast, poorly understood "non-coding" regions harbour the majority of mutations linked to human disease. To dissect their function, Google Deepmind developed AlphaGenome, a model that predicts functional genomic measurements from DNA sequences with unprecedented breadth and resolution. This capability was unlocked by training on massive, multimodal genomic datasets, allowing the model to learn the intricate regulatory "grammar" of the genome. As the most comprehensive model of gene regulation to date, AlphaGenome achieves state-of-the-art results, consistently outperforming even specialised single-modality approaches across a wide range of variant effect prediction tasks.

AI success starts long before models and algorithms. This session focuses on what it truly takes to build AI-ready data and digital foundations in biotech R&D. From data quality and standardization to connected systems and intelligent automation, speakers will explore how organizations can prepare their labs for trusted, scalable AI and move from experimentation to real operational value.

13:20 - 13:40: Building the Lab-in-the-Loop Foundation for AI - Patrick Collins, VP of Automation and Scientific Operations, Relation

Drug discovery is a learning problem. We do not yet know enough about biology and disease, nor do we have sufficient high-quality data to enable the de novo generation of novel therapeutics. Unlocking the full potential of AI and ML in this setting therefore requires a lab-in-the-loop approach in which learning systems are tightly connected to real-world data generators capable of probing biological space in disease-relevant and increasingly complex systems. Multimodal patient-derived data provides a powerful foundation for linking causal biology to clinical phenotype, with perturbational omics and translational cellular models providing insights into disease biology. Achieving this at scale requires the deliberate integration of software, hardware, and human expertise coupled to a laboratory automation strategy that strikes the right balance of experimental diversity, flexibility, and scale. Such integration is essential for generating the fit-for-purpose datasets required for modern AI/ML in target discovery, translational biology, and drug discovery.

13:45 - 14:05: How Integrating Data, in Silico Modeling, and Automation Accelerates Innovation and Scale-up for Cultivated Meat - Gautier Koscielny, VP Digital Biology, Ivy Farm Technologies

Ivy Farm is pioneering a scalable, sustainable, and economically viable approach to cultivated meat production. This presentation will showcase how we have built an integrated digital ecosystem to accelerate both R&D and scale-up. It will also demonstrate how digital and AI/ML capabilities strengthen scientific pipelines through experimental reproducibility and data integrity, support regulatory readiness, and enable rapid, continuous innovation cycles.

14:10 - 14:35: Panel: Building the AI-ready foundation, moderated by Wollis Vas, Customer Success Manager, Benchling

                    - Jake Galson, Vice President of Technology, Alchemab
                    - Henry Farmery, Head of Operations, ForthTX

                    - Ashwin Nandakumar, Founder, Granza Bio

General-purpose AI models are rapidly closing the gap on domain-specific tools, scientists are building production pipelines overnight with AI coding assistants, and yet most biotech teams still struggle with the unglamorous work of structuring their data. In this panel, leaders from three companies at different stages; Alchemab Therapeutics, Granza Bio, and Forth Therapeutics share hard-won lessons on what "AI-readiness" actually looks like in practice. They'll debate whether bespoke models still justify the investment when off-the-shelf alternatives keep improving, how to navigate the tension between moving fast and capturing structured data, and where AI is genuinely accelerating R&D versus where it's creating new risks.

Scaling AI is both a technology and a transformation challenge. This session looks at how biotech companies can successfully evolve their AI capabilities from early-stage innovation to enterprise-grade operations. Through expert discussion and roundtable exchange, participants will explore the tools, governance models, and change management practices needed to ensure AI delivers long-term scientific and business impact at scale.

15:05 - 15:25 - From Research to Routine: Making Models Stick - Peter Quicke, Research Software Lead, Hoxton Farms

As organisations scale, deploying machine learning models becomes as much a cultural and operational challenge as a technical one. In this talk, we’ll examine the real-world barriers to turning models into dependable decision-making tools. Moving from notebook analyses to automated pipelines powering high-throughput screens, we’ll share practical lessons on how to transform a promising model into something scientists genuinely trust and use.

15:30 - 15:55 - Panel: Scaling with Intention, moderated by Nunzia Sposito, Customer Success Manager, Benchling

                    - Ania Wilczynska, Senior Director Bioinformatics & AI, bit.bio

                    - Vincenzo Di Cerbo, Program Head of Digital innovation, Cell and Gene Therapy Catapult

This session brings together Ania Wilczynska, Senior Director of Bioinformatics & AI at bit.bio, who has been in the thick of it — navigating what it takes to move from early AI wins to organisation-wide adoption inside a fast-growing biotech, and Vincenzo Di Cerbo, Programme Head of Digital Innovation at Cell and Gene Therapy Catapult, who sits at the intersection of industry, innovation and policy, working across the sector to help life sciences organisations build the digital and AI foundations they need to scale. Together, they bring a rare combination of inside-out and outside-in perspectives on what it really takes to move from AI ambition to operational reality.

16:00 - 16:20 - High-Quality Data as the Foundation for AI-Driven Discovery - Vaibhav Bhardwaj, Scientific Solution Consultant, Benchling 

AI breakthroughs like AlphaFold2 prove that powerful models require high-quality, structured data. Yet, as discovery becomes fully automated, the volume of data from high-throughput screens is outpacing our ability to process it. This session addresses the "dual automation" challenge: physical sample handling and digital data capture. We will explore how structured platforms and the Model Context Protocol (MCP) are transforming scientific workflows—enabling complex, cross-system analyses in minutes rather than hours.