Accelerating drug discovery with AI: Insights from Isomorphic Labs

Modeling Complexity

The exploration of machine learning models in drug discovery reveals the complexity and potential of these technologies. Max Jaderberg highlights the challenges of modeling protein data due to the limited scale compared to NLP or computer vision, emphasizing the need for integrating physical, biological, and chemical priors into data representation 1. He also discusses the emergence of global models like AlphaFold, which generalize beyond their training data to predict novel protein sequences and molecular interactions 2. This approach allows for a versatile platform that can be adapted to various drug design programs.

The big difference, Lucas, is that the scale of data is many orders of magnitude lower than in something like NLP or computer vision.

--- Max Jaderberg

These insights underscore the importance of developing models that can handle diverse and complex biological data.

   

Data Transformation

Transforming biological and chemical data for machine learning models is a critical step in drug discovery. Max Jaderberg explains the use of transformers and the ongoing challenge of finding optimal ways to process complex data structures like proteins and genomes 3. He notes that while some data can be treated as linear sequences, the intricate structures of biological molecules often require more sophisticated representation strategies. Sergei Yakneen adds that understanding the underlying biology is crucial for effective data representation.

There's massive amounts of structure that are not linear in the genome.

--- Sergei Yakneen

This highlights the need for innovative approaches to accurately capture the nuances of biological data in AI models.