Collaborative Research: SCH: Integrated Analysis of Single-Cell and Spatially Resolved Omics Data

What genes are expressed in a cell shapes a cell's function and behavior. The set of genes that are expressed in a cell at a given time is highly regulated and can be impacted by many factors, including the cell's spatial context. Consequently, a fundamental question for understanding the function and behavior of cells in both health and disease is how such gene expression levels may be different in healthy versus diseased cells.

Recent technological advances, including single-cell RNA-seq and spatially resolved omics, now allow us to measure what genes are expressed in what cells simultaneously for hundreds to thousands of genes and thousands to millions of cells while preserving the spatial arrangement of cells within the original tissue. Such technological advances and the resulting data generated present the opportunity to develop a deeper understanding of how different patterns of gene expression define different cell-types or cell-states, how these different cell-types and cell-states are organized within tissues, and how gene expression and spatial organizational patterns may be different in health versus diseased tissues.

However, as these datasets are huge and complex, new analytical tools are required to extract relevant biological insights in an interpretable manner. This project aims to develop and test these tools, with a particular emphasis on building simple interpretations to facilitate potential clinical translations to cost-effective procedures. The PIs plan to integrate the research with education, train students, especially young women, by involving them in the research, and share the developed curricula materials and computer code with the public.

Cellular identity and heterogeneity are shaped by a multitude of intrinsic molecular and spatial-contextual factors. Profiling cells and their molecular states within their native spatial context provides a modality to connect cellular organization and function. Recent technological advances are enabling such genome-wide molecular profiling of the transcriptome and proteome in a spatially resolved manner at single-cell and near-single-cell resolution.

However, new statistical methods and computational tools are still needed to model and analyze these high-dimensional spatially resolved omics measurements in order to extract relevant biological insights in an interpretable manner. This research aims to deduce from the abundant information that these data modalities provide, a compressed, but interpretable, summary, based on minimal sets of genes that capture the underlying cellular transcriptional heterogeneity.

The scientific approach is based on statistical spatial modeling and information theory, designing algorithms for approximating the entropy of groups of random variables, and using integer programming methods to search for small subsets of variables that represent best the whole collection initially observed. The research development and validation will be based on multiple datasets, some of them being publicly available and some shared by collaborators at Weill-Cornell Medicine.

These datasets will involve three of the main spatially resolved transcriptomic modalities recently introduced (VISIUM, CODEX, MERFISH), and will contain traditional imaging modalities using immunohistochemistry as well. Understanding how some of these modalities can be used to enhance partial observation in other modalities is also part of the research program.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.