Genome-wide profiling of the hierarchical control of autophagy dynamics using deep learning

Recycling of nutrients through autophagy is a crucial mechanism for cells to sustain internal stability in a fluctuating environment. Dysregulation of the process has been associated with a range of human disorders, and the core components of autophagy have been comprehensively outlined. However, limited insight into its systems-wide dynamical control has hampered predictive modeling and effective in vivo manipulation. Here, we mapped yeast genomic influences on autophagy dynamics in response to changes in nitrogen levels. Using time-resolved high-content imaging coupled with deep learning, we examined the kinetics of autophagy activation and inactivation in 5919 gene deletion mutants, and classified their profiles based on temporal responsiveness and activation potential, as well as their contribution to autophagosome formation and clearance. By integrating these profiles with functional and genetic network data, we unveiled a hierarchical and multi-layered control of autophagy dynamics and exposed novel regulatory features of the core components and well-established nutrient-sensing pathways. Furthermore, by leveraging multi-omics resources and explainable machine learning to model genetic perturbation effects, we identified the retrograde pathway as a central time-varying transcriptional modulator of autophagy execution. These findings offer valuable insights into the systems-wide tuning of autophagy, and advance our understanding of the dynamical control by providing genome-wide quantitative data under concurrent genetic and environmental interventions. We further anticipate that our study can serve as a blueprint for high-content, deep learning-driven exploration of complex dynamical processes in any organism. ### Competing Interest Statement The authors have declared no competing interest.