Flexible open-source automation for robotic bioengineering

Emma J. Chory, Dana Gretton, Erika A. DeBenedictis, Kevin M. EsveltbioRxiv 2020 · DOI

TL;DR

This paper introduces Pyhamilton, an open-source Python package that unlocks the full potential of liquid-handling robots for complex bioengineering tasks. By overcoming the limitations of traditional, narrowly designed software, Pyhamilton enables novel experiments, accelerates high-throughput assays, and facilitates sophisticated feedback-controlled bioprocesses, ultimately democratizing advanced robotic automation in research labs.

What Problem Does This Paper Address?

Liquid-handling robots are indispensable tools in modern biotechnology, automating repetitive and laborious tasks. However, the software controlling these robots is often rigid, designed to mimic basic hand-pipetting. This inflexibility prevents researchers from fully leveraging the robots' advanced capabilities, such as executing complex liquid transfer patterns, implementing real-time feedback control, or integrating with other lab equipment. This gap limits the scope of experiments that can be automated, hindering progress in areas like biomanufacturing, synthetic biology, and fundamental biological research. The authors aim to bridge this gap by providing a flexible, open-source software solution that empowers researchers to push the boundaries of robotic automation.

Methods

The core of this work is the development of Pyhamilton, a Python package that acts as an intermediary between users and Hamilton liquid-handling robots (STAR and STARlet models). Key methodological contributions include:

  • Python-based Programming: Pyhamilton allows users to program robots using standard Python, inheriting its powerful paradigms like object-oriented programming, exception handling, and version control. This makes robotic method development more accessible and robust.
  • Abstraction Layers: The package provides Python classes that abstract robotic actions (e.g., aspirate, dispense) and consumables (e.g., plates, pipette_tips), simplifying the creation of complex protocols.
  • Simulation Capabilities: Pyhamilton integrates with Hamilton's run control software to simulate methods, enabling experimental planning and error handling before execution on the physical robot.
  • Asynchronous Programming: The framework supports asynchronous tasks, allowing multiple operations (like plate reading and pipetting) to occur concurrently, significantly boosting throughput.
  • Feedback Control Integration: Pyhamilton facilitates the implementation of feedback-controlled loops, enabling robots to adjust actions in real-time based on measurements from integrated instruments (e.g., plate readers).
  • Modular Design: The software is designed to be modular, allowing for the integration of auxiliary instrument modules and enabling its use as a standalone package or as part of a larger project.

Datasets

The paper does not describe the use of pre-existing datasets for training or evaluation in the traditional machine learning sense. Instead, the authors demonstrate Pyhamilton's capabilities through a series of experimental applications:

  • Automated Bacterial Assays: Used to demonstrate increased throughput.
  • Population Dynamics Simulations: Employed to showcase complex pipetting patterns.
  • Feedback-Controlled Bacterial Cultures (Turbidostats): Used to maintain hundreds of cultures in log-phase growth.
  • Metabolic Profiling of Protein Production: A large-scale experiment involving nearly 500 continuous cultures to explore nutrient fitness landscapes.

These applications serve as benchmarks and demonstrations of Pyhamilton's functionality, rather than datasets in the conventional sense.

Key Insights and Ideas

  • Democratizing Advanced Robotics: The most significant insight is that by leveraging open-source Python and established software engineering principles, complex and previously inaccessible robotic automation can be made available to a wider range of researchers.
  • Beyond Hand-Pipetting Mimicry: Traditional robotic software is often limited by its focus on replicating manual pipetting. Pyhamilton's true power lies in enabling entirely new experimental paradigms that are impossible or impractical to perform manually.
  • The Power of Asynchronicity: Implementing asynchronous programming is crucial for achieving high-throughput in complex, multi-step biological experiments, allowing for simultaneous data acquisition and manipulation.
  • Real-time Feedback is Game-Changing: Integrating real-time monitoring and feedback control transforms robots from simple liquid dispensers into dynamic experimental platforms, essential for maintaining delicate biological systems like turbidostats.
  • Bridging Computer Science and Biology: The paper exemplifies a growing trend of applying robust computer science methodologies to biological automation, leading to more powerful and flexible experimental tools.

Key Results

  • Doubled Assay Speed: Pyhamilton was used to double the speed of automated bacterial assays by enabling simultaneous pipetting into two 24-well plates.
  • Complex Pipetting Patterns: The software successfully executed complex pipetting patterns to simulate population dynamics, including bounded and unbounded flow across simulated geographic barriers.
  • High-Throughput Turbidostat Maintenance: Pyhamilton enabled the real-time monitoring and feedback-controlled maintenance of nearly 500 bacterial turbidostats, achieving stable log-phase growth for extended periods.
  • Metabolic Fitness Landscape Mapping: The system was used to comprehensively map the carbon, nitrogen, and phosphorus fitness landscape for recombinant protein production in E. coli, analyzing hundreds of conditions simultaneously.
  • No Measurable Cross-Contamination: In the turbidostat experiments, a robust tip-cleaning and individual-tip-per-culture strategy ensured no measurable cross-contamination between wells.

Limitations and Future Work

  • Hardware Specificity: While Pyhamilton aims for abstraction, its current implementation is primarily focused on Hamilton STAR and STARlet liquid-handling robots. Expanding compatibility to other robot platforms would increase its reach.
  • Learning Curve: Although Python-based, mastering the full capabilities of Pyhamilton and integrating it with diverse lab equipment may still require a learning investment from users.
  • Future Work: The authors suggest that Pyhamilton can be extended to support a wider array of robotic actions and auxiliary instruments. They also envision its application in areas such as genetic knockout screening, small-molecule compound libraries, and studying microbiome system dynamics, highlighting the vast potential for future research and development. The open-source nature of the project inherently encourages community-driven expansion and application.

Summary updated

Abstract

INTRODUCTION Liquid handling robots have become a biotechnology staple 1,2 , allowing laborious or repetitive protocols to be executed in high-throughput. However, software narrowly designed to automate traditional hand-pipetting protocols often struggles to harness the full capabilities of robotic manipulation. Here we present Pyhamilton, an open-source Python package that eliminates these constraints, enabling experiments that could never be done by hand. We used Pyhamilton to double the speed of automated bacterial assays over current software and execute complex pipetting patterns to simulate population dynamics. Next, we incorporated feedback-control to maintain hundreds of remotely monitored bacterial cultures in log-phase growth without user intervention. Finally, we applied these capabilities to comprehensively optimize bioreactor protein production by maintaining and monitoring fluorescent protein expression of nearly 500 different continuous cultures to explore the carbon, nitrogen, and phosphorus fitness landscape. Our results demonstrate Pyhamilton’s empowerment of existing hardware to new applications ranging from biomanufacturing to fundamental biology.
BibTeX
@misc{chory2020flexible,
  author = {Emma J. Chory and Dana Gretton and Erika A. DeBenedictis and Kevin M. Esvelt},
  title = {{Flexible open-source automation for robotic bioengineering}},
  year = {2020},
  doi = {10.1101/2020.04.14.041368},
  howpublished = {bioRxiv},
}

Loading related content...

Figure