Abstract
Cells suppress protein synthesis during starvation, but how translation recovers upon nutrient repletion remains poorly understood. To address this gap, we followed translational recovery in budding yeast after repletion of glucose, nitrogen, or phosphate, combining time-resolved metabolomics, structural proteomics, transcriptomics, and spike-in-calibrated ribosome profiling. Nutrient addition rapidly altered ribosome occupancy in distinct, transcript-specific programs. To identify their sequence determinants, we developed RiboNet, a deep learning framework that predicts ribosome load and positions along transcripts from mRNA sequence. RiboNet-guided in silico mutagenesis revealed reduced pausing at serine codons upon glucose repletion and transiently increased pausing at leucine codons following nitrogen repletion. These responses coincided with altered cognate tRNA charging and metabolite-associated structural changes in aminoacyl-tRNA synthetases. RiboNet also identified a conserved 5′ UTR motif, GAUY, that increased reporter protein output following glucose repletion. Our findings reveal how metabolic state alters the contributions of mRNA sequence features to translation initiation and elongation.