Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

Bioinformatics, Computational and Systems Biology

Bioinformatics, Computational and Systems Biology - Poster Session A

Poster A4 - AI-Assisted Lab Automation Leveraging Traditional Instruments and Open Source Hardware in Jupyter Notebook

Thursday, October 24, 2024
10:00 AM - 11:00 AM EST
Location: Exhibit Hall E, F & G

Introduction:  Automated systems, such as liquid handling robots and data analysis platforms, enable researchers to conduct high-throughput assays with minimal human intervention, reducing potential errors while improving reproducibility. These ‘experiment-ready’ systems are often expensive and render previously purchased equipment obsolete, hindering their adoption by many researchers. Many electronic lab notebooks (ELN) and schedulers are limited in an all-in-one capability to design experiments, run protocols, collect all data, and report results. Furthermore, chatbots and AI could be used to fabricate experiments and fake research, raising skepticism about the widespread use of AI in science. Herein, we propose a collaborative framework where technicians, AI, robots, and traditional wet lab instruments work together to conduct bioscience experiments.

For this development version, our Jupyter Notebook is organized in accordance with the SciNote ELN structure. Tasks consist of code cells and are grouped such that a safe stop is possible between chronologically organized tasks. Experimental design is a collaboration between humans and chatbots, as is instrument programming for hardware modules required to execute the method. Method execution is a collaboration between the technician operating both traditional and automated instruments, with images and video recordings as proof of execution. All data, images, and videos are pulled into the notebook for storage, visualization, and reporting in HTML or other formats. With orchestration in Jupyter Notebook, we can ensure unified data and protocol management to facilitate cost-effective, transparent, and efficient automation adoption.

Materials and

Methods:  For a simple use case to test the workflow’s functionality, we chose to record and analyze BSL-1 E. coli growth in a microplate reader and microscope. Labware and reagent labels were imaged and cataloged with Pytesseract for the sake of recordkeeping, as is typical for handwritten lab notebooks. Sterile media was manually prepared under surveillance by camera imaging and video recording. A video-recorded Opentrons OT-2 liquid handling robot loaded two 96-well plates with E. coli in LB plus sterile LB control wells to monitor for contamination. One plate was used for 600nm absorbance measurements under 37oC incubation in a Tecan Infinite M200 microplate reader. Incubated microscope imaging occurred at 37oC for each well of the second plate. All output images, data, and metadata were uploaded to a single AWS S3 Bucket. BASIN-ML was used to analyze the micrographs, and a custom Python script was used to visualize the growth curves generated by the microplate reader.

Results, Conclusions, and Discussions: Here, we developed a lab automation workflow packaged in Jupyter Notebook. Instruments and software were treated as “modules,” each with standalone functionality and an ability to be orchestrated with the Jupyter Notebook. These modules include an Opentrons OT-2 robot with a programmable HEPA fan, a JuLI Stage incubator microscope, a Tecan microplate reader, a microcontroller with DS18B20 temperature sensors, a programmable label maker, an inventory camera, a camera-monitored mass balance and hot-stir plate, AWS S3 Buckets, and BASIN-ML. Close module proximity was required due to a 15-foot limit on USB signal strength. Digital readouts were important for data logging using USB webcams and Pytesseract for optical character recognition. These cameras, alongside the wirelessly connected Opentrons, are controlled by the Client laptop, which also contains the Jupyter Notebook. The S3 Bucket retrieves data from modules that cannot be connected to the Client via USB or wireless signal. They also store micrographs to prevent overloading the Client’s local hard drive.

Our first test case involved simple E. coli growth observance in well plates. We chose this simple method to test the collaborative ability of technicians, AI, automation hardware, and traditional instruments to work together and produce reproducible, transparent results without sacrificing practicability. Reproducibility is quantifiable by analyzing growth curves and micrographs across multiple runs. Our results show the feasibility of ethically incorporating AI/ML for experimental design, image analysis, and experiment reporting while showing evidence that a real wet lab method was executed.

This approach leverages an inexpensive, collaborative automation workflow that is friendly to existing lab instruments yet has a view toward AI-assisted robot science. It prioritizes maximum data collection to ensure experiment transparency and compliance with FAIR Guiding Principles for data management. It also serves as a bulwark against forged research by using real-time camera footage and retaining all instrument inputs and outputs. Ongoing work involves integrating more AI/ML and automation to collaborate with researchers for accelerated design, redesign, execution, analysis, and publication of original research.

Acknowledgements (Optional):  Funding was provided by the Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health (P20GM103443). Initial lab automation system development was funded by NSF EPSCoR RII Track 1 (#1849206) and Track 2 (#1920954).