Automated Design of Experiment Optimisation

How machine learning can enable complex design of experiments (DOE), making multi-factorial media optimisation simple, efficient and robust.


Your Challenge

Biological systems are complex networks of often individually simple biochemical reactions; it is these networks of reactions which give rise to the emergent properties that govern modern biology. Despite this complexity, many scientists still optimise these systems with a traditional reductionist “one factor at a time” approach. 

In contrast, Design of Experiments (DOE) is a much more powerful and efficient multi-factorial approach. However, despite the utility of DOE methods they are notoriously hard to execute in the laboratory, requiring weeks of planning and numerous error-prone liquid handling steps. Automation should be a good solution, but historically automated DOE solutions are limited to a fixed number of factors, or to legacy design setups that may not be suitable for executing future DOEs. Re purposing such inflexible pre-programmed systems for altered or new cases requires considerable programming experience, time and effort. 

Synthace's Solution

To allow biologists to robustly question the inherent complexity of biology, we must move to ways of working that are built around the generation, structuring and analysis of large multi-factorial data sets. To enable this we developed Antha, an easy-to-use biological operating system that can implement multi-factorial DOE optimisation for you. Antha allows you to leverage all the benefits of multi-factorial optimisation without the burden of extensive planning, or manual liquid handling.[1]

Antha works with existing laboratory hardware, and due to its easy to use interface, can execute workflows as simple or as complex as the user desires. Furthermore, protocols can easily be reproduced and shared across teams allowing for multiple identical experiments to be run simultaneously.



Synthace's software platform accelerates the whole DOE workflow – from planning through to scaling. The eloquent control of liquid handling is at the core of the Antha platform, resulting in a biology-centric user experience that improves reproducibility and traceability, all whilst enabling the scientist to undertake more complex experimentation.

BG #ededed.jpg

Case Study:  Complex Growth Medium Liquid Handling Automation

Our client, a multinational pharmaceutical company, was experiencing a poor yield of a key protein expressed in E. coli and therefore conducted an Antha enabled DOE optimisation of growth media conditions. The client utilised the Antha DOE integration to easily upload their designs. The experiment was then simulated, and error checked via the drag and drop Antha workflow editor, before being automated on a Gilson PIPETMAX® controlled by the Antha platform.

By using our platform the pharmaceutical company was able to rapidly complete a 384-run DOE study using 74% fewer tips and liquid handling actions. Crucially as Antha can directly integrate experimental design files from widely used DOE software, such as JMP and Design Expert, it seamlessly integrated into the company’s existing workflow. The planning time for the DOE optimisation was also reduced from 1-2 weeks to 2 hours and the robot running time reduced by 20%. By using Antha the pharmaceutical company was able to test more conditions simultaneously, more accurately, and thus establish an optimal combination of nitrogen sources, buffer compositions and carbon sources that resulted in 300% increase in protein titre.


Ready to Revolutionise your Workflow?

The example given here illustrates a proof of concept case for integrating Computer Aided Biology into your DOE workflow. With applications rapidly coming on-line for advanced bio processing solutions, construct assembly, assay development, and drivers for more pieces of lab hardware the opportunities for transforming your lab’s productivity with digitally enabled DOE are endless.


Book a Demo

BG #ededed.jpg

[1] Sadowski, Michael I. et al., Trends in Biotechnology , Volume 34, Issue 3 , 214 - 227

[return to top]