Compression of Structural, Cartographic, and Multimodal Cell Data

To develop data compression algorithms for Human Cell Atlas data that support the eventual end use of the atlas, whether in-place execution of network analysis, machine learning, or visualization.

The Varshney lab developed theoretical, practical, and computational insights into the nature of data and algorithms that are needed for downstream analysis tasks for the Human Cell Atlas. They developed a novel geometric technique for imputing data in support of Non-negative Matrix Factorization (NMF) algorithms by extending the idea of optimal recovery from numerical analysis and signal processing to the conical geometry in this setting.