We have used our GeneGPS® platform to develop codon-optimization algorithms for a range of hosts. Open circles in the graphs below show the expression of sets of genes designed and tested as training sets. We use machine learning to identify design properties that are associated with highly expressed genes, and then incorporate these into a back-translation algorithm. The filled circles show the performance of genes designed using the resulting algorithm. We are always interested in customized GeneGPS® algorithms to optmize genes for new expression hosts.