Evozyne participation at NVIDIA GTC conference

Evozyne co-founder and Computational Lead Andrew Ferguson presented on AI acceleration in drug discovery at NVIDIA GTC last week.

?GTC is a leading conference on artificial intelligence, computer graphics, and data science, which highlights recent hardware and software developments.

?Ferguson was invited to speak on the GTC panel titled, "AI-Accelerated Molecular Dynamics and Protein Engineering for Drug Discovery," which was moderated by Anthony Costa, Senior Alliance Manager, NVIDIA, and Venkatesh Mysore, Principal Solutions Architect, NVIDIA.

?Other distinguished panelists included Juan Carlos Mobarec, Associate Director, AstraZeneca; Mark Moraes, Head of Engineering, D. E. Shaw Research; Huafeng Xu, CTO, Roivant Discovery; Virginia Burger, co-founder and CEO, New Equilibrium BioSciences; and Andrew Ban, Vice President, Computing, Arzeda.

?The panel focused on how GPU hardware and AI algorithms are influencing and enabling drug discovery, protein engineering, and molecular modeling, the role for tight coupling between AI and high-throughput experimentation, and the next frontiers.

?The panel discussed exciting developments in AI-learned force fields and the accuracy and cost trade-offs relative to classical parameterizations of the interaction potentials. The importance of community and internal benchmarks for testing and validation of novel AI approaches resonated strongly with the participants: the deluge of new techniques and architectures emerging in this fast-moving field makes this a priority in the triage and prototyping of new methodologies.

?The panelists also discussed whether the culture of open-source sharing of codes has lowered the barrier to entry in the field of AI-enabled molecular modeling and design but emphasized the paramount importance of tight coupling to experiment to improve and refine the computational models and provide the ultimate validation of the computational predictions. Within this vein, an innovation the panel found particularly exciting was the potential for tighter coupling between computation and experimentation within the paradigm of "self-driving laboratories" that have the potential to realize large accelerations in molecular discovery pipelines.

?Overall, the panel conveyed a high degree of optimism about the great strides in AI-enabled molecular modeling and design enabled by GPU hardware and algorithmic advances.?Panelists identified innovative language models, cloud computing, and coupling with high-throughput experimentation as a potentially fruitful route to efficient realization of clinical candidates.