Abstract
the practice of engineering biology now depends on the ad hoc reuse of genetic elements whose precise activities vary across changing contexts. methods are lacking for researchers to affordably coordinate the quantification and analysis of part performance across varied environments, as needed to identify, evaluate and improve problematic part types. We developed an easy-to-use analysis of variance (AnoVA) framework for quantifying the performance of genetic elements. For proof of concept, we assembled and analyzed combinations of prokaryotic transcription and translation initiation elements in Escherichia coli. We determined how estimation of part activity relates to the number of unique element combinations tested, and we show how to estimate expected ensemble-wide part activity from just one or two measurements. We propose a new statistic, biomolecular part 'quality', for tracking quantitative variation in part performance across changing contexts. Genetic engineers must specify a priori the precise activities of biomolecular parts for use in integrated synthetic systems 1-11. Improvements in methods and tools for synthesizing and assembling DNA 12,13 additionally challenge practitioners to design genetic sequences that result in precise expression of hundreds of coding sequences 14-16. Meanwhile, distributed communities of researchers struggle to collectively assemble, measure, validate and share collections of standard biological parts 17,18. Additionally, sophisticated biotechnology applications addressing medical or environmental needs demand improved tools for reliably estimating the expected performance of engineered systems 19. Against this backdrop of needs, studies of biological part activities 20-24 have revealed that the quantitative activity of genetically encoded elements is often highly context dependent 15,25-29. For example, engineers and biologists have generated and studied libraries of synthetic expression control elements 21,30-38 on an ad hoc basis and across relatively limited contexts 14,39. In some cases, researchers have used first-principle models to develop predictors of element function that attempt to account for context impacts 30,36,37,40-42. Though valuable, these models cannot fully capture the impact of changing contexts on genetic element function. More recently, researchers have developed passive and active genetic insulators such that the functioning of one element might not corrupt a neighboring element 43-45. Yet, with the lack of systematic and quantitative data detailing how and to what extent different types of genetic elements interact, it remains unclear that such projects have focused on regularizing the most difficult element-element junctions or leveraged the simplest normalizing molecular mechanisms. We developed an easy-to-deploy mathematical framework that can be used to score the intrinsic activities of genetic elements to track how such activities vary (or not) across changing contexts. We propose that variation in part activity can serve as a quantitative score of part quality to concisely summarize the reliability of part reuse. For example, we might score a 'promoter' element that never initiates transcription in any and all contexts as a 'high-quality promoter encoding zero activity' , whereas an element that in some contexts initiates transcription and in others does not, would be a 'low-quality promoter encoding intermediate activity'. Genetic element 'quality' in these two examples captures the extent to which users of elements can rely on the reported behaviors; a promoter that is known to never initiate transcription would be of particular value for establishing negative controls used in quantifying both transcription promoters and terminators. To develop and demonstrate the method, we constructed a full combinatorial library of frequently used transcription and translation elements in E. coli, expressed two genes at two temperatures