Team:Marburg/Labautomation

L A B A U T O M A T I O N


Introduction to Robot and AI-Powered automation subgroup

This year at iGEM Marburg, we decided to incorporate the AI and robot revolution into our cutting-edge Synthetic Biology research. With the Opentrons OT-2 pipetting robot that we won last year and advanced machine learning methods, our objective was to design automation protocols that accelerate research in Synthetic Biology. This was realized in two ways, namely by delegating mindless works to pipetting robots and by achieving a high degree of reproducibility. The latter being a point that has been a common complain in biology and can be mitigated by standardization, which is very welcomed by the community (for more details see Integrated Human Practices). We envision a future of Synthetic Biology, where people exchange robot protocols on top of text description of an experiment to ensure a high degree of reproducibility. We believe that our approach of infusing the best of computing and robotic world into Synthetic Biology will elevate the field even further.

As a proof-of-concept, we concentrated on the hugely popular Golden Gate cloning method; more specifically on some crucial parts of it such as plasmid purification and colony picking. These aspects currently pose a major challenge for a comprehensive automation of the cloning process and therefore create a great demand from the community.

In our plasmid purification project, we designed a workflow that enables the utilization of Promega Wizard® MagneSil® plasmid purification product with the Opentrons OT-2. For the successful implementation of this step, it was essential to overcome the challenge to incorporate products from different manufacturers, such as Opentrons Magnetic Module and QInstrument D-30T elm shaker into our protocol. We achieved this by designing and printing custom-made hardware adapters to guarantee stable and reproducible labware. We have also made it our aim to make OT-2 protocols more dynamic, which allows easy customization even for non-programmers. Specifically, it allowed us to create a flexible protocol that is up- and down-scalable i.e. with more or less samples amount ranging from 1 to 48 per run. All of which are fully automated without human intervention in-between steps.

Colony picking is one of the toughest steps to automate due to its complex nature. After we have tested several methods such as hough circle transformation from OpenCV for the initial colony detection we discovered that the problem is even harder than we’ve originally thought. To address this issue, we talked to several experts from the field of labautomation and computer vision, which led us to the decision to implement the colony detection using a convolutional neural network method. It was clear to us that we needed a method that was highly flexible, could run on as many machines as possible and that would ensure a fast and accurate detection of the naturally small colonies. After several weeks of research, we decided to use the state-of-the-art Faster R-CNN method which has little computing cost and is available through the well supported, open source TensorFlow machine learning framework. As with any AI approach, the quality and quantity of the training data is of the utmost importance. We overcame the data collection issue with the help of the incredible iGEM community via a colony picture competition. The competition was a big success, where we collected more than 300 valuable pictures that were used for the training of our algorithm. Finally, the colony detected by the AI were translated into robot movement via an in-house algorithm conceived by our team.

Lastly, our team designed a Graphical User Interface (GUI) that allows biologists to easily design their own golden gate protocol for the OT-2. We understand that if we want to bridge the gap between forefront biology research and engineering, we have to make it intuitive for every technical lay person. That is why we created the GUI which will also help with our vision of standardization.

In summary, the automation lab has helped elevating iGEM Marburg to another level by infusing high precision method of advanced robotics, mathematics, and machine learning into the workflow of our biology team. Our vision is to create a new trend of protocol exchange to kick start a world of biology research with the highest degree of reproducibility. As a proof-of-concept for our vision, we addressed some existing bottlenecks in biological workflows such as colony picking or plasmid purification using cutting edge approaches from fields outside of biology and thus exhibiting the highest level of interdisciplinary research. Lastly aligned with the iGEM philosophy, all of our work will be published as open-source upon which other research labs can build. Our code and 3D designs will be available in a Github repository, which will be publicly available online. We envision that our protocols and approaches to be a pioneering work for a world of highly reproducible Synthetic Biology experiment.


Placeholder image
Automation overview : plating, picking, prepping