(228cw) Comparing the Predictive Accuracy of Translation Rate Models Across 16,685 mRNA Sequences

Precise control over protein expression is required when engineering biological systems such as genetic circuits or metabolic pathways for chemical production. Modifying the ribosomal binding sites (RBS) of the encoding mRNA is one of the most commonly employed and effective strategies used for tuning protein expression levels. Early on, biological engineers achieved this by using previously characterized RBS parts or designing ad hoc RBS sequences, both of which often resulted in unpredictable protein expression in various genetic contexts. To address this challenge, a number of models of translation initiation have recently been published including the RBS Calculator (versions 1.0, 1.1, 2.0), UTR Designer, RBS Designer, and EMOPEC. These models attempt to map the complex relationship between sequence and expression, but do so with varying biophysical descriptions of the interactions between the 30S ribosome and the mRNA. These differences reveal the ongoing debate over what the definitive determinants are that control translation initiation. Here, we present a systematic model comparison on a diverse 16,685 mRNA sequence-expression database, and the result of that analysis, a definitive list of determinants that regulate translation initiation. We used a combination of sequence structure categorization, model error analysis, and RNA folding dynamics simulations to identify that the rate and efficiency of mRNA folding into low free energy structures can modulate the rate of translation initiation by over 1,000-fold â?? a phenomena not properly accounted for in any of the existing models. With these new insights, we look to develop a more accurate biophysical model of translation initiation to enable more rapid and efficient design of genetic systems.