Biochemistry's Hidden Matrix: Atomic Constellations
Our latest paper details how we extract informative "constellations" of interacting atoms within these structures, yielding over 11.5 million built-for-purpose data points today. This also establishes a path to an estimated 500+ million data points in the short-term and potentially 2.3x1014 from the Tree of Life proteome long-term.
By combining public data with our own proprietary Cryo-EM data, as well as rigorous benchmarking and iterative refinement, we promote continual data quality improvements. Our analyses also confirm optimized distributions for model training.
All of this scales up chemical space coverage and opens doors for our AI models to reach unprecedented performance levels in generating novel therapeutic insights - ultimately leading to drugs in the market.
We encourage you to dive deeper into our data pipeline and approach within the full paper - “Biochemistry's Hidden Matrix: Atomic Constellations”.
This paper is part of a series that reviews various aspects of Model Medicines GALILEO™ AI Drug Discovery Platform and its two parallel drug discovery modules CHEMPrint™ and Constellation™.
Each paper demonstrates the capabilities of the GALILEO™ platform through quantitative case studies. The first Constellation™ paper presented our proprietary, cutting edge Cryo-EM data acquisition pipeline.
LEAD AUTHOR
Navya Ramesh, Machine Learning Engineer
SUPPORTING AUTHOR
Bastiaan Bergman, Ph.D., VP of Engineering
