Computational Studies of the Patterning Systems in Early Drosophila Embryos

Abstract

In development, the body plan for an organism must be executed properly for its survival. This is accomplished by a collaborative effort by genes that form gene regulatory networks (GRNs) that have several redundancies built in to ensure development remains robust. The early embryonic development of fruit flies (Drosophila melanogaster) is an exceptional model system to study GRNs that control the timing and spatial extent of gene expression. This work focuses on unraveling the mechanisms of robustness in patterning along the two major axes of symmetries in the early embryo - dorsal-ventral (DV) and anterior-posterior (AP). Using the computational tools of model development, network analysis, image analysis, and machine learning, the GRNs in early Drosophila embryonic development are investigated. Along the DV axis, model development uncovered novel mechanisms that keep development robust when the dosage of genes is perturbed. Next, the feedforward loop network motif was analyzed, generally and in DV patterning, and its general features were extracted that contribute to robustness. Along the AP axis, an image analysis pipeline was developed to analyze gene expression domains in sagittal planes of the embryo with high accuracy. The reliability of the quantitative gene expression data obtained was significantly enhanced by building in multiple levels of fallbacks and using geometric data about the embryo's shape. Further, a machine learning model was developed to stage the embryos in early development to the precision of a few minutes using nuclear morphology instead of relying on gene expression. Finally, a function optimization algorithm was developed to estimate the free parameters of systems biology models, which generally have several free parameters, and each parameter is poorly constrained. The algorithm, which extends an evolutionary strategy approach, converges to better solutions with higher probabilities over successive runs.