Bio-design automation (BDA) is a new discipline that aims to automate the synthetic biology pipeline using workflows adapted from electronic design automation (EDA) (Rodrigo and Jaramillo 2013; Chen 2009). Through the adoption of BDA, the ‘Design-Build-Test’ cycle can be fully automated, with only the initial ‘Specification’ and final ‘Learn’ stages requiring human input. In the future, this could even be automated with the use of AI and machine learning. BDA can be split into two different categories; the automated hardware such as the integration of robotics, and the software that allows for in silico optimisation through computational modelling (Densmore and Bhatia 2014).

Automated hardware such as liquid handling robots and microfluidic systems allow for significantly reproducible, standardised workflows, while removing human error as a potential source of variation (Beal et al. 2016). Integrated software for synthetic biology allows for in silico modelling enabling the use of the available databanks. From this, complex genetic circuits can be built, designed and tested through computer simulation, without ever having to enter the laboratory for wet-lab experimentation. For an exhaustive review on the range of software available see Appleton and colleagues (2017) work. These synthetic biology workflows require data standards for BDA to be successfully implemented. Part standards must encode more than just the sequence, requiring the inclusion of additional data such as environmental and experimental information whilst also providing computational models of construct behaviour and measurements of performance (Galdzicki et al. 2014). Synthetic Biology Open Language (SBOL) is a data standard aiming to tackle this challenge. Developed by the synthetic biology community, it allows the exchange of biological circuit design in a format that can be understood by software and repositories (Galdzicki et al. 2014). Successful integration of SBOL into the wider scientific community would allow for a more directly integrated design workflow and increase the reproducibility of experimental results (Beal et al. 2012; Peccoud et al. 2011).

Laboratory robotics and smart lab equipment provides the platform for high reproducibility in experiments, however due to their current price bracket are unavailable to the majority of labs. This has driven recent developments in making affordable technologies for lab equipment, notably in the release of the Opentrons OT-2 liquid handling robot and the smart incubator by Incuvers Incorporated. With these developments and the ever-growing library of software, the likelihood that BDA will be integrated into most modern labs is high.

Following calibrations, two transformed colonies for each test device and both controls were used to inoculate LB medium containing chloramphenicol (CAM) and incubated overnight at 37 °C with shaking at 220 rpm. Overnight cultures were diluted 1:10 and the OD600 adjusted to 0.02 with LB with CAM to a final volume of 12 ml. Fluorescence and Abs600 were taken at 0h and 6 hours of incubation at 37 °C with 220 rpm shaking. Test devices, plasmid backbone and protocol workflow are shown in figure 2.

All test devices produced growth to OD600 reading in excess of 0.3, except test device (TD) 4. Despite lower growth than other transformants, TD4 produced the highest mean fluorescence reading of 79.1 a.u., as was expected as the strongest promoter of the Anderson collection (parts.igem.org/Promoters/Catalog/Anderson). Figure 3A and 3B show the colony 1 and colony 2 Abs600 values for the controls and each test device at 0 hours and 6 hours respectively.

The relatively poor growth and high fluorescence levels effectively cancelled each other out when readings were converted to Fluorescence per OD600 and MEFL per OD600 measurements, resulting in TD4 transformants producing the highest expression levels (figure 3.1). The high fluorescence and MEFL per OD600 reading for TD4 despite lowest growth suggests expression of TD4 is not fully representative of the relative promoter strength; expression levels are interdependent with growth rate, with higher growth rates expected to produce higher expression levels (Scott et al. 2010). Expected fluorescence levels based on relative promoter strength reported for the Anderson collection of promoters did not match entirely the results produced here. In particular, TD5 utilising the promoter J23104 was expected to be the second strongest but yielded only the fourth highest fluorescence reading of 22.14 and 21.42 for colonies 1 and 2 respectively. Similarly, the highest fluorescence reading was recorded by TD1 (expected strength: third), though this test device produced the widest range in fluorescence reading between the two colonies (r = 36.2), despite both colonies having the closest OD600 reading of any of the test devices (r = 0.007). In addition to the iGEM repository documentation for relative strengths of the Anderson promoter collection, previous literature has also demonstrated that the J23101 and J23104 promoters should have almost equal strength (He et al. 2017).

While TD5 appeared to underperform compared to promoter activity previously reported in the literature, subsequent sequencing of test devices revealed that colonies labelled as TD5 had in fact been transformed with the positive control device. This may have simply been the result of human error when pipetting or labelling over the process of the study. The consistence of TD5 underperformance across multiple replications of the study suggests that this occurred early on. The variation in expression visible is particularly alarming for the J23101 promoter, which has been proposed and utilised in the literature as a reference promoter to characterise relative strengths of other promoters as relative promoter units (Kelly et al. 2009).

While the InterLab study is an effective way of gathering large sets of data surrounding the parts and protocols, it does not consider all sources of variation within datasets or the variability between data sets. As there are many factors which cause variability in microbial protein expression and productivity, some of these factors may be more favoured than others, leading to an overrepresentation of these in the metadata.