Theses Doctoral

Deciphering complex diseases by engineered tissues and multimodal data integration

He, Siyu

Complex diseases are multifactorial, driven by genomic and environmental perturbations that cause cellular and organ dysfunctions. Cells in diseased organs are strongly coupled with their surroundings. Thus, uncovering the interactions between various cells provides insight into how they adapt to the environment, respond to stimuli, and progress to malignancy. Current technologies in engineered tissues, high-resolution images, single-cell analysis, and spatially resolved transcriptomics have shown the capability to model diseases and characterize morphologic and genomic features with desired resolutions.

Meanwhile, machine learning is emerging as a powerful tool for analyzing these large-scale biomedical data and unraveling the complexity of disease mechanisms. However, the capability of engineered tissues in modeling cell- cell interaction, thus leading to disease malignancy, remains unknown due to the challenges of multi-modal biomedical data integration.I explored intercellular interactions through engineered tissues and computational models in this dissertation.

I first studied two diseases by modeling the vascular structure and functions through blood vessel organoids, namely 22q11.2 Deletion syndrome and Proteus syndrome, which are associated with vascular disorders. The integration of imaging and single-cell transcriptomic analysis unravels the disease mechanisms relating to the dysfunction of fibroblasts secreting extracellular matrix and affected endothelial-mural cell interaction. I further extended the study of cell-cell interaction to spatially-resolved transcriptomics.

By developing an end-to-end deep generative model with the integration of spatial transcriptomics and histology, I identified that spatial heterogeneity in the tumor microenvironment and different cell interactions distinguishes metaplastic breast cancers from other triple-negative breast cancers. Moreover, I investigated the integration of theoretical physics models, single-cell transcriptomic data, and statistical machine learning to interpret the interaction between mesenchymal stem cells and cancer cells for potential strategies of cancer cell therapy.


  • thumnail for HE_columbia_0054D_17858.pdf HE_columbia_0054D_17858.pdf application/pdf 8.1 MB Download File

More About This Work

Academic Units
Biomedical Engineering
Thesis Advisors
Leong, Kam W.
Azizi, Elham
Ph.D., Columbia University
Published Here
June 14, 2023