Bowlin, Peter, Li, Zihao, Fouad, Anthony, Du, Angelica, Kassouni, Alexander, Bhirgoo, Priya, Teng, Christopher, Fang-Yen, Christopher
[
International Worm Meeting,
2021]
The time and labor required for worm picking is a major bottleneck for many C. elegans experiments, especially those requiring a large number of strains. Many genetic screens and genetic manipulations would benefit greatly from an automated method for worm pushing. We developed a robot capable of transferring worms between agar plates using movements similar to those used for manual worm picking. The robot contains a motorized 3D stage that positions a wire loop pick mounted on a robotic arm to manipulate worms on an array of standard plates. Capacitive touch sensing is used to monitor contact between the pick and the agar substrate and provide feedback for the fine movements needed for picking. We constructed a dual-magnification fluorescence and bright field microscope capable of identifying developmental, morphological, or fluorescence based phenotypes of individual worms at high resolution while simultaneously imaging the entire plate at low resolution. We developed software to identify, classify, and track worms using a combination of machine vision methods, including motion detection, adaptive thresholding, and a convolutional neural network trained to recognize worms. In a test of its fluorescence-based sorting capabilities, the robot accurately identified, phenotyped, picked, and transferred worms to other plates at a rate of about 3 animals per minute. We are developing a high-level scripting language that will enable the robot to autonomously perform multi-step procedures, such as integrating extrachromosal arrays, performing genetic crosses, generating clonal populations of mutagenized worms, and other tasks. Automation of worm manipulation will both increase researchers' productivity and enable experiments that are impractical using standard methods.