%0 Journal Article %A Iman Farasat %A Manish Kushwaha %A Jason Collens %A Michael Easterbrook %A Matthew Guido %A Howard M. Salis %T Computational design to efficiently search, map, and optimize multi-protein genetic systems in gram-negative and gram-positive bacteria %D 2013 %R 10.1101/001008 %J bioRxiv %P 001008 %X Engineering multi-protein genetic systems to maximize their performance remains a combinatorial challenge, particularly when measurement throughput is limited. We have developed a computational design and modeling approach to build predictive models and identify optimal expression levels, while circumventing combinatorial explosion. Maximally informative genetic system variants are first designed by the RBS Library Calculator, an algorithm that optimizes the smallest ribosome binding site library to efficiently search the expression space across a >10,000-fold range with tailored search resolutions, sequence constraints, and well-predicted translation rates. We validated the algorithm’s predictions using a 644 sequence data-set, within single and multi-protein genetic systems, modifying plasmids and genomes, and in Escherichia coli and Bacillus subtilis. We then combined the search algorithm with kinetic modeling to map the mechanistic relationship between sequence, expression, and overall activity for a 3-enzyme biosynthesis pathway, requiring only 73 measurements to forward design highly productive pathway variants. The combination of sequence desi gn and systems modeling accelerates the optimization of many-protein systems, and allow previous measurements to quantitatively inform future designs. %U https://www.biorxiv.org/content/biorxiv/early/2013/12/02/001008.full.pdf