Engineering biology and automation-Replicability as a design principle

Abstract

Applications in engineering biology increasingly share the need to run operations on very large numbers of biological samples. This is a direct consequence of the application of good engineering practices, the limited predictive power of current computational models and the desire to investigate very large design spaces in order to solve the hard, important problems the discipline promises to solve. Automation has been proposed as a key component for running large numbers of operations on biological samples. This is because it is strongly associated with higher throughput, and with higher replicability (thanks to the reduction of human input). The authors focus on replicability and make the point that, far from being an additional burden for automation efforts, replicability should be considered central to the design of the automated pipelines processing biological samples at scale-as trialled in biofoundries. There cannot be successful automation without effective error control. Design principles for an IT infrastructure that supports replicability are presented. Finally, the authors conclude with some perspectives regarding the evolution of automation in engineering biology. In particular, they speculate that the integration of hardware and software will show rapid progress, and offer users a degree of control and abstraction of the robotic infrastructure on a level significantly greater than experienced today.

Keywords: automation; bioinformatics; synthetic biology.

FIGURE 1

Design space in engineering biology. The design space is the set of all combinations of the independent variables. It can be split into a subset with all the variations in the genetic design, and another including experimental conditions and protocols (parameters and sequence of actions).

FIGURE 2

Implementation of the DBTL cycle in biofoundries. The DBTL cycle uses an expansion‐reduction strategy. The expansion portion (build and test phases) enables the controlled introduction of variations into the system, and does so in decoupled stages, to yield a combinatorial effect. The reduction portion (learn phase) identifies its most promising regions, so they can be targeted in later rounds.

FIGURE 3

Automated tasks and automated pipeline. (1–Top): Every automated task has two components—a software layer, coupled to a mechanised layer. Instructions for the task are communicated to the mechanised layer by the software layer. (2–Middle): Multiple tasks are connected into an integrated automated pipeline. Information flows are denoted by black arrows, and flows of physical objects by white arrows. There is now a chaperoning IT infrastructure in charge of collecting all the generated data and storing them. (3–Bottom): the IT infrastructure needed to support replicability contains an orchestrator, charged with sending the instructions to the various automated platforms, monitoring the advancement of the workflow, and validating the content of the information passed between components.