Techniques for Spatial Analysis of C. elegans Anatomy and Reporter Expression

Quantitative analysis of microscopy images is integral to investigating biological phenomena. Despite a variety of tools to aid in analyzing C. elegans images, quantitative microscopy studies are still difficult due to the flexible and deformable nature of the nematode. These differences in posture and shape must first be corrected before analysis. Manual approaches to solve these problems are time intensive and infeasible for large datasets. Additionally, current automated tools rely on high-magnification imaging using labeled nuclei as fixed markers for comparison. Labelling can be achieved either with transgenic animals or fluorescent dyes; however, both of these can be impractical for some studies. Thus, there is a need for a more generalized method for quantitatively analyzing C. elegans images without requiring transgenic animals or fluorescent dyes. To address this need, I have developed a set of automated, machine-learning based tools to locate and align anatomical landmarks in brightfield images of C. elegans. This allows for effective correction of positional and anatomical differences among individuals and over time. A key challenge in this work was identifying the best representation of anatomical landmark positions with which to train the models. Image-to-image regression proved most successful in this application. The toolkit described herein can be applied to many imaging modalities (fluorescence, brightfield, etc.) as long as there is a corresponding brightfield image. I used these methods to examine population variation in anatomy, to explore morphological changes over time, and to analyze temporal and inter-individual trends in reporter fluorescence. The work presented in this dissertation provides the foundation for a generalized image analysis toolkit that can be used by the C. elegans community in studying a variety of biological questions. In the second part of this dissertation, I outline additional uses for which this toolset could be employed in the future. The tools are easy to use, train, and extend, and are publicly available as an open-source Python package on GitHub