Project summary We propose to use quantitative, high-throughput methods to understand the consequences of variation in translation elongation on gene expression. New methods including ribosome profiling provide a genome-wide view of the motion of ribosomes along transcripts. We recently developed a neural network based on ribosome profiling data that captures information about what makes a ribosome move faster or slower, and then used that same information to design synonymous sequences that are not just decoded at different rates but also actually make more or less protein. This raises two questions that we propose to address here: how does slow decoding result in diminished protein expression, and what consequences does this rate variation have in vivo? First, we will expand our preliminary results from yeast, adapting our neural network model to understand the impact of variation in translation elongation in mammalian systems. We will measure changes in translation elongation in different cell types and with differential activity of translation elongation factors. Second, we will investigate and model the interplay between translation initiation and translation elongation rate to understand how translation elongation can be rate-limiting for protein production. We will also identify trans-acting factors that modulate this effect, using a genome wide CRISPR screen. Third, we will develop a more complete neural network model relating gene sequence to ultimate translation output, incorporating not just local sequence context but also positional effects and other factors. This proposal presents new experimental systems that can quickly and sensitively measure the consequences of codon choice and identify factors affecting how different codons determine translation output.