Mingze Dong

Generated by GPT 4o.
I am a CBB (Computational Biology & Bioinformatics) PhD student at Yale University, fortunate to be co-advised by Prof. Yuval Kluger and Prof. Rong Fan. Before my PhD, I earned a BS degree in Integrated Science from Peking University, focusing on applied math and biology. I also conducted research in the Lin lab and van Dijk lab.
My research interests lie in the foundations of machine learning (identifiability, scaling laws, etc.) and how to design better methods through deep consideration of these fundamental principles, particularly for biological applications. My earlier works include applications for single-cell perturbation modeling (CINEMA-OT, Nature Methods 2023), graph noise metrics predicting GNN performance gain (ESNR, ICML 2023), and inference of causally interacting genes from spatial omics data (GEASS, ICLR 2023).
In a recent line of work, I develop identifiability theories and practical implementations to infer intrinsic/spatial variations in spatial omics data (SIMVI) and perform zero-shot disentanglement of biological states through pretraining on massive single-cell compendiums, overcoming the well-known batch effect problem (scShift). I pursue novel insights, theoretical rigor, and empirically superior performance together in my work. Additionally, I am engaged in extensive collaborations with biologists to analyze real-world biological data (Nature 2023, with more publications in progress).
Feel free to contact me if you are interested in potential collaboration or just a chat!
selected publications
- Nature MethodsCausal identification of single-cell experimental perturbation effects with CINEMA-OTNature Methods, 2023
- In submissionScaling deep identifiable models enables zero-shot characterization of single-cell biological statesbioRxiv, 2024
- Nature Commun
Editor’s HighlightSIMVI disentangles intrinsic and spatial-induced cellular states in spatial omics dataNature Communications (Featured Article), 2025 - Under revision
- ICMLTowards Understanding and Reducing Graph Structural Noise for GNNsProceedings of the 40th International Conference on Machine Learning, 2023