What Makes a Cat a Cat: The Application of Deep Learning to the Automated Detection of Crystals

Automating Image Analysis with Deep Neural Networks

Elucidating protein structure lies at the heart of any therapeutics discovery program; otherwise, it’s like designing a key without knowing the shape of its lock. For decades, crystallography has been the tool of choice for studying macromolecule structure, yet today parts of this critical process are still cripplingly labor-intensive.

Automating the slowest step in the process, checking if crystals have grown, has been a topic of research for decades, but many hurdles still remain. New advances in image analysis, powered by deep neural networks, can provide a way forward.

This white paper discusses:

  • The challenges of automating crystallography research
  • The impact of neural networks on image analysis
  • Future outlooks for research in the space