Oliver May delivered a presentation on SynSilico, a joint venture aimed at integrating AI capabilities with the pharmaceutical industry. He began by acknowledging the growing interest in machine learning and briefly explained SynSilico's mission to enhance AI applications in pharma. May emphasised that while AI is transformative, it is crucial to combine it with human understanding of biological systems.
SynSilico was formed through the collaboration of AI experts and the DSM pharma development unit. May highlighted the importance of the team, particularly those who manage computer systems. He explained that the company relies on both machine predictions and human validation to ensure accuracy.
The initial focus of SynSilico was on optimising pharma product development processes using Bayesian optimisation. This approach was also applied to a wide variety of other fields, even bread baking. May mentioned that these tools are now available as APIs for various applications. In the insulin discovery field, SynSilico uses AlphaFold and other models to screen binding modes of substrates to proteins. However, May noted that binding modes alone do not predict catalysis, leading the company to introduce molecular dynamics capabilities for more accurate predictions.
May discussed the development of an enzyme discovery toolbox, which included protein folding, active site identification, and docking. He acknowledged the limitations of current tools and stressed the need for high-quality predictions. To address this, SynSilico leveraged its protein engineering capabilities to design single-chain antibodies using generative AI.
The platform developed by SynSilico allows for rapid design and production of antibodies, which can be used for various pharmaceutical applications. May highlighted the economic and clinical feasibility of camelid-based antibodies, which were chosen for their ability to recognise diverse epitopes. He concluded by emphasising that SynSilico's approach combined AI with rational design, leveraging both in silico predictions and biochemical expertise to achieve success.
