Computational topology of Drosophila melanogaster embryos: Emergent anatomy from the spatiotemporal distributions of hemocytes.

Abstract

Hemocytes migrate within Drosophila melanogaster embryos during their morphogenesis. The spatial distribution of the hemocytes reflects the emergent anatomy of the embryos. It poses a significant challenge to statistical characterization using conventional methods, such as mean-square displacements of tracks and Ripley's L function. This challenge arises from the complex heterogeneity of hemocyte dynamics and the variability of the organism's anatomy due to inherent nonlinear stochastic processes. We apply topological data analysis (TDA) to three different genotypes of Drosophila embryos: wild type, LanB1^{Df} (extracellular matrix mutants), and SCAR^{Δ37} (motility mutants). Computationally efficient methods are developed to characterize the embryos' topologies. Based on the Wasserstein distances, the LanB1 mutant is topologically distinct from the wild type and SCAR mutant, which are indistinguishable. Furthermore, an increase in the Betti-1 number (the number of holes in the hemocyte configurations) is observed across all samples with age, which is consistent with prior findings on cellular tiling behavior due to contact inhibition of locomotion. We also apply TDA methods to the hemocyte tracks within the embryos, allowing the wild type and SCAR mutant to be distinguished. Computational topology thus provides a robust method to quantify the emergent anatomy of embryos that is well suited to the complex heterogeneous datasets observed with hemocyte dynamics.