Leveraging metabolic modeling and machine learning to uncover modulators of quiescence depth

Abstract

Quiescence, a temporary withdrawal from the cell cycle, plays a key role in tissue homeostasis and regeneration. Quiescence is increasingly viewed as a continuum between shallow and deep quiescence, reflecting different potentials to proliferate. The depth of quiescence is altered in a range of diseases and during aging. Here, we leveraged genome-scale metabolic modeling to define the metabolic and epigenetic changes that take place with quiescence deepening. We discovered contrasting changes in lipid catabolism and anabolism, and diverging trends in histone methylation and acetylation. We then built a multi-cell-type machine learning model that accurately predicts quiescence depth in diverse biological contexts. Using both machine-learning and genome-scale flux simulations, we performed high-throughput screening of chemical and genetic modulators of quiescence and identified novel small molecule and genetic modulators with relevance to cancer and aging.