posted on 2018-01-24, 00:00authored byThomas Decoene, Gert Peters, Sofie L. De Maeseneire, Marjan De Mey
Fine-tuning
biosynthetic pathways is crucial for the development
of economic feasible microbial cell factories. Therefore, the use
of computational models able to predictably design regulatory sequences
for pathway engineering proves to be a valuable tool, especially for
modifying genes at the translational level. In this study we developed
a computational approach for the de novo design of
5′-untranslated regions (5′UTRs) in Saccharomyces
cerevisiae with a predictive outcome on translation initiation
rate. On the basis of existing data, a partial least-squares (PLS)
regression model was trained and showed good performance on predicting
protein abundances of an independent test set. This model was further
used for the construction of a “yUTR calculator” that
can design 5′UTR sequences with a diverse range of desired
translation efficiencies. The predictive power of our yUTR calculator
was confirmed in vivo by different representative
case studies. As such, these results show the great potential of data
driven approaches for reliable pathway engineering in S. cerevisiae.