Soliciting Feedback from the community

As I mentioned previously on github, we’ve been having a number of conversations within the DeepChem development team about the future of DeepChem. We want to open out our discussion to the broader community to get your feedback on what would bring most value:

  1. Do you currently use DeepChem? For what? If you used it in the past, why did you stop? Or if you considered using it at some point, why did you decide not to?
  2. What are the most useful and least useful aspects of DeepChem?
  3. What’s the biggest pain-point of DeepChem?
  4. How could DeepChem be improved to make it more relevant to your interests?
  5. What’s your primary deep learning framework? TensorFlow, PyTorch, or something else? If DeepChem supported an alternative framework like Pytorch Geometric (for example), would that make a difference for you?

Please leave your comments below! Alternatively, if you’d prefer to answer these questions anonymously, you can use this google form or answer on github. Thank you for taking the time to help us figure out a good path forward for the project :slight_smile:

Over the next week or two, we’ll start proposing new directions for the DeepChem project based on the community feedback we’ve received. We hope to see lively discussion of what people would like to see DeepChem go!

2 Likes

Excited to see how DeepChem shapes up, and how I can contribute more to it, in the coming months!

  • Do you currently use DeepChem? For what?

We are starting to use DeepChem to generate continuous embeddings for small molecule similarity search. ( We tried previously pretrained mol2vec embeddings and compared to a traditional Tanimoto metric. )

  • How could DeepChem be improved to make it more relevant to your interests?

There is a growing body of work that is using GANs/Reinforcement Learning to navigate large chemical spaces. Might be interesting to include that in DeepChem.

  • What’s your primary deep learning framework? TensorFlow, PyTorch, or something else?

TensorFlow and PyTorch.

1 Like