Harnessing the Predictive Power of Microtiter Plate Models for Multiple Products in Fermentation Scale Up

Amyris is a global renewable products company providing sustainable alternatives to a variety of non-renewable resources.  We generate and screen >15,000 microbial strains each week using random mutagenesis and rational strain engineering.  At this scale we currently rely on 96-well microtiter plates (MTPs) for cell culture and high-throughput assessments of titer, cellular rates, and strain health.  We developed a multivariate model that combines these separate measures of strain performance into a single metric that predicts strains' ability to produce farnesene, a terpene Amyris converts into fuels and emollients.  Results from the MTP stage are a leading indicator for bioreactor performance; the correlation between MTP data and various metrics of bioreactor performance, including yield and productivity, currently exceeds 0.85 (as presented at SIMB 2014).  Based on the farnesene MTP process we have developed a set of universal MTP culture models and assays that are compatible with a broad panel of molecules.  With this new platform, the time to optimize new methods fell from 12-24 weeks to less than 6 weeks.  Such efficiency in method development allows resources to be devoted to novel plate models and measurements that may add more predictive power to our multivariate model and further increase our MTP to tank correlation.  Using this platform of universal plate models and assays, Amyris has improved its R&D efficiency and scaled many molecules from MTPs to bioreactors in less time while running separate strain improvement pipelines for the development of multiple products in parallel.