Abstract
Comparative analyses of natural sequences or variant libraries are often used to infer mechanisms of expression, activity and evolution. Contingent selective histories and small sample sizes can profoundly bias such approaches. Both limitations can be lifted using precise design of large-scale DNA synthesis. Here, we precisely design 5 E. coli genomes worth of synthetic DNA to untangle the relative contributions of 8 interlaced sequence properties described independently as major determinants of translation in Escherichia coli. To expose hierarchical effects, we engineer an inducible translational coupling device enabling epigenetic disruption of mRNA secondary structures. We find that properties commonly believed to modulate translation generally explain less than a third of the variation in protein production. We describe dominant effects of mRNA structures over codon composition on both initiation and elongation, and previously uncharacterized relationships among factors controlling translation. These results advance our understanding of translation efficiency and expose critical design challenges.