This research introduces UNAGI, a deep generative model designed to analyze cellular dynamics and facilitate in silico drug discovery for complex diseases like idiopathic pulmonary fibrosis and COVID-19. By combining a graph VAE-GAN architecture with gene regulatory network inference, the tool maps how cell populations transition across different stages of a disease. UNAGI distinguishes itself from previous methods by using an iterative training strategy that emphasizes disease-specific markers, improving the precision of cell embeddings. The model enables researchers to perform simulated perturbations, virtually testing thousands of compounds to identify those that might shift diseased cells back toward a healthy state. Validated through precision-cut lung slices, this framework identifies potential therapeutic candidates, such as nifedipine, without requiring prior experimental drug-response data. Ultimately, UNAGI offers a scalable, unsupervised solution for decoding the temporal complexity of multifactorial disorders.

References:

• Zheng Y, Schupp J C, Adams T, et al. A deep generative model for deciphering cellular dynamics and in silico drug discovery in complex diseases[J]. Nature Biomedical Engineering, 2025: 1-26.