
A new artificial intelligence (AI) model called UNAGI may offer hope for patients suffering from idiopathic pulmonary fibrosis (IPF) a deadly, complex lung disease with no cure.
Developed by an international team of researchers from Yale School of Medicine, KU Leuven (Belgium), and McGill University. UNAGI is capable of learning disease behavior and autonomously suggesting therapeutic options.
How UNAGI Works
UNAGI is a deep generative neural network designed to revolutionize treatment for lung diseases like idiopathic pulmonary fibrosis (IPF). By analyzing gene expression data from over 230,000 lung cells, it identifies disease-specific markers and fibrosis-related pathways. For instance, it flagged nifedipine a hypertension drug as a potential IPF treatment. This AI-driven approach accelerates biomedical discovery and offers new hope for rare diseases with limited therapies.
Nifedipine Among Promising Therapies Identified
UNAGI flagged 8 possible treatments for idiopathic pulmonary fibrosis (IPF). These included HDAC inhibitors apicidin and belinostat, which demonstrated anti-fibrotic potential, but were not tested in clinical trials that featured people with IPF.
However, UNAGI also flagged nifedipine, a standard drug for treating hypertension which has an excellent safety profile, as a potential treatment. Three separate laboratory tests on human lung tissue that were modeled for IPF showed that nifedipine inhibited scar formation.
Why IPF is a Good Fit for AI
Idiopathic Pulmonary Fibrosis (IPF) presents with uneven lung scarring, unpredictable progression, and multiple comorbidities. Since the disease affects different lung regions differently, researchers use tissue from transplant patients to simulate various stages.
Conclusion
UNAGI represents a breakthrough in precision medicine for IPF, using AI to model disease progression and uncover targeted treatment options. By linking gene expression patterns to existing drugs like nifedipine, it offers a faster, more efficient path to new therapies for complex diseases.