Abstract:
Automated quantification of microscopy datasets plays an important role in understanding biological phenomena. This requires robust computational methods for accurate object detection, tracking, and segmentation. In this thesis I present the development of image processing pipelines to quantify nematode embryo viscosity, spindle dynamics and cell cycle progression in evolutionary related nematode species, dynamics of single mictotubule filaments driven by molecular motors, bacterial shape allometry and RBC adhesion from microscopy datasets. Spindle dynamics, observed during the first mitotic division, show large variations across nematode species. To investigate these differences, we first interrogated the rheological properties of the embryo cytoplasm using a DIC object tracking software (DICOT). By combining granule microrheology with force estimates acting on the mitotic spindle, quantified through spindle laser ablation, we reveal a trade-off between cytoplasmic viscosity and spindle forces on the appearance of spindle oscillations during the first anaphase stage. Additionally, our quantification revealed a strong correlation between cellular viscosity and cell division time. Motivated by this, we developed a Deep Learning based model to automatically reconstruct the progression of the first mitotic division in diverse Caenorhabditis embryos from label-free DIC time series. The classifier predictions rely on stage-dependent morphological patterns, and in the future, the tool can be used to study cell stage-dependent changes in intracellular rheological properties. The mechanics of spindle oscillations are driven by complex interactions between MT and molecular motors. These interactions have been studied using in vitro re- constitution systems. To understand the mechanics of MT-driven spindle oscillations, our lab previously devised an in vitro system to reproduce single filament oscillations. However, quantification of filament geometry using existing tools proved to be limiting. To address these limitations, we developed a segmentation pipeline, named as KnotResolver, to accurately trace intersecting or branched filaments that enables the quantification of oscillation frequencies. Morphology of cells shows high heterogeneity in allometric scaling of surface area and volume. To understand this scaling relationship, we developed a cell morphology analysis pipeline that can be applied to multiple imaging modalities. In preliminary work, we have also developed a pipeline to quantify RBC cytoadhesion in in vitro microfluidic chambers. Overall, the computational image-analysis pipelines provide valuable insights into various cellular biological processes, facilitating further advancements in the field.
