Health Science Reviews

By: Simon Mufara

Imagine if your grandma’s blood pressure meds could also treat Alzheimer’s. Or if a cancer drug could unexpectedly ease Parkinson’s symptoms. This isn’t science fiction; it’s called drug repositioning, and it’s changing the future of medicine. Developing a new drug from scratch can cost more than R47 billion and take over a decade. Meanwhile, thousands of approved drugs sit on shelves, often overlooked.

What if we could match these drugs with new diseases using AI?

In this study, researchers from Xiamen University and the Cleveland Clinic introduced deepDR, an artificial intelligence framework that learns from massive drug networks to predict new drug–disease combinations. Think of it like Spotify, but instead of recommending music, it suggests treatments.

How does it work?

deepDR utilizes deep learning to integrate information from 10 massive biological networks, including drug targets, side effects, chemical similarities, and gene functions. First, it processes these networks through a multi-modal deep
autoencoder, which simplifies and fuses data into a compact “fingerprint” of each drug. Then, a variational autoencoder takes these fingerprints and predicts what diseases the drug might also treat.

What did they find?

deepDR outperformed traditional machine learning models like SVMs and Random Forests, showing over 90% accuracy in cross-validation tests. Even better when tested against real clinical trials data, it correctly predicted drugs for Alzheimer’s and Parkinson’s that were already under investigation, including risperidone and methylphenidate.

Why does this matter?

As someone working on a breast cancer drug response prediction model for African patients, I found deepDR inspiring. It shows how AI can help us repurpose drugs faster and smarter. With frameworks like deepDR, we can build scalable tools that cut costs, reduce side effects, and bring precision medicine to communities, especially those that need it most.

Reference
Zeng, X., Zhu, S., Liu, X., Zhou, Y., Nussinov, R., & Cheng, F. (2019). deepDR: a
network-based deep learning approach to in silico drug repositioning. Bioinformatics,
35(24), 5191–5198. https://doi.org/10.1093/bioinformatics/btz418