DeepChem 2025 Targets

Here are some early musings on goals I’d like us to hit as a community in 2025:

  • Swapping to all Torch defaults for models
  • Deprecating all TensorFlow/Jax code formally
  • First equivariant models (SE(3) transformers, E3NN style models): Adding equivariance to DeepChem has been tricky. I want us to mature this engineering effort and get it into DeepChem core
  • First RNA-seq/ATAC-seq models: Single cell deep learning has been booming. I’d like to see us get some models into

Suggest more below!