I've been working on QUBE Predict, a cloud platform for drug response prediction built around graph-based biological descriptor learning.
Rather than relying solely on end-to-end models, the framework focuses on reusable biological descriptor representations that can be paired with conventional machine learning algorithms.
The platform includes:
• REST API
• Web interface
• Batch prediction
• Authentication and API keys
• Cloud deployment
The methodology has been described in a publicly available technical paper.
Paper:
https://doi.org/10.5281/zenodo.21455765
Live Demo:
https://qube-predict.streamlit.app/
GitHub:
https://github.com/gulfamh1999-afk/qube-predict-ui
I'd really appreciate technical feedback, especially on the representation-learning methodology and platform architecture.