Matrix Factorization-based Technique for Drug Repurposing Predictions
Problem
Developing new drugs is hugely costly and slow, with about 85% of proposed molecules failing the first three FDA approval phases, so finding new indications for already approved drugs is needed.
Why it matters
High drug development costs and failure rates delay access to treatments for patients.
What the AI did
An enhanced Non-negative Matrix Tri-Factorization machine learning method uses shortest-path evaluation of drug-protein pairs on a protein-protein interaction network to infer novel drug indications and previously unconsidered protein targets.
Results
not stated in source
Impact
The method enables prediction of novel indications and new protein targets for approved drugs.
Record
| Venue | IEEE Journal of Biomedical and Health Informatics |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Gaia Ceddia, Pietro Pinoli, Stefano Ceri, Marco Masseroli |
| Citations | 56 (OpenAlex, at import time) |
| References | 54 |
| Model or method | Non-negative Matrix Tri-Factorization (NMTF) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1109/jbhi.2020.2991763 |
| Open access | green |
| Keywords | drug repurposing, matrix factorization, machine learning, protein targets, drug-protein pairs |
Importance rating
major — millions of lives, an industry or a field
Abstract quantifies that ~85% of new proposed molecules fail the first three FDA phases, making drug development hugely costly and slow; repurposing approved drugs is framed as crucial for patient access to treatments.