Anish Virdi

Undergraduate Researcher @Michigan Medicine

Ann Arbor, MI, US

Project Description: During mitosis, unaligned chromosomes scaffold the production of a protein called the Mitotic Checkpoint Complex (MCC); this protein inhibits anaphase, giving chromosomes time to align. Quantifying the rate at which chromosomes produce MCC is an open challenge because (1) the protein is short-lived and (2) this rate depends on dynamic intracellular conditions. I am using systems biology to infer, not directly measure, per-chromosome MCC production rates-Using high-throughput fluorescence microscopy to study anaphase inhibition in live cells across expansive degrees of freedom- Designed an algorithm to detect unaligned chromosomes in low-resolution images: https://github.com/anishjv/uchrom_cycb- Using both systems of ordinary and stochastic differential equations to model anaphase inhibition and infer MCC production- Using light-sheet microscopy to fine-tune the rate quantifications derived from my high-throughput studies.Project Description: Deep learning-based cell-segmentation models are typically trained on hand-annotated data, which limits both dataset size and model performance. I developed Cell-APP, a tool that automates annotation in transmitted-light microscopy, enabling researchers to train custom cell-segmentation models. https://github.com/anishjv/cell-AAP-Devised and built Cell-APP’s Python 3 software, which automatically segments and classifies cells in microscopy images- Deployed Cell-APP to train cell-segmentation models that allow our lab to analyze > cells per week- Wrote two manuscripts and built various graphical user interfaces to support other labs’ use of Cell-APP- By building Cell-APP, I reduced training dataset compilation time by ~90% as compared to hand-annotation.