BioNTech (BNTX) AI Day 2024 summary

Strategic Vision and AI Integration

• BioNTech and InstaDeep are building an AI-first, personalized immunotherapy platform, leveraging genomics, immunology, and advanced AI to accelerate individualized treatment development and drug discovery.

• The partnership has evolved from early collaborations to InstaDeep's acquisition, embedding AI across BioNTech's platforms and functions.

• InstaDeep and BioNTech have developed a near exascale supercomputing cluster, Kyber, in Paris, with over 224 NVIDIA H100 GPUs and 86,000 CPUs, enabling large-scale AI model training and research.

• AI is embedded across the R&D pipeline, from histology and genomics to proteomics and protein design, with demonstrated improvements in speed, accuracy, and scalability.

• InstaDeep maintains a dual strategy: driving innovation for BioNTech’s pipeline while also serving external clients and open-source collaborations, ensuring broad impact and sustained AI leadership.

Key Technological Advancements

• Introduction of Bayesian Flow Networks (BFN), a novel generative AI model for unified modeling across protein sequence, structure, and function, powering models like ProtBFN and AbBFN.

• Launch of AbBFNX, a multimodal antibody model enabling flexible, conditional generation and optimization of antibody sequences for rapid therapeutic design.

• DeepChain platform integrates generative models, nucleotide transformers, and AI assistants, supporting both internal and external users for accelerated R&D and experiment design.

• InstaNovo, a de novo peptide sequencing tool, increases peptide identification rates by 40% in immunopeptidomics, expanding target discovery for cancer immunotherapies.

• Foundation models for genomics, like Nucleotide Transformer and SegmentNT, deliver state-of-the-art performance in genome annotation, splicing, and regulatory element prediction.

Business Model and Future Outlook

• InstaDeep’s supercomputing and AI infrastructure supports BioNTech’s proprietary projects and external partnerships, focusing on scalability, cost efficiency, and flexibility.

• AI-derived molecules and constructs are evaluated alongside expert-designed ones, with the best results from collaboration between domain experts and AI researchers.

• The platform’s open-source and collaborative approach allows external academics and biotech companies to fine-tune models while maintaining proprietary data security.

• AI is already used in clinical applications, such as RNA-encoded antibodies and cytokines, with human trials underway and promising therapeutic translation levels achieved.

• Ongoing expansion of AI applications into manufacturing, clinical trial optimization, and supply chain, with a pragmatic approach to leveraging both internal and external solutions.