Development of tools for analyzing cell-cell communication using spatial transcriptomic data

PROJECT SUMMARY Biological tissues, consisting of different cells, specialize in a group of processes and functions through coordinated activities of many cells. Cell-cell communication (CCC) by ligand-receptor interaction provides a major mechanism for such coordination. Until recently, dissecting CCC required perturbations of selected genes or proteins regulated within a specific CCC link, presenting major challenges for experimental approaches. Single-cell genomics that profiles genes and their activities at individual cell level provides an unprecedented opportunity for systematic screening of all potential CCC links among cells. During the past three years, various computational tools, including ours, have enabled CCC inference and analysis using the nonspatial single-cell (sc) RNA-seq data, leading to many important biological discoveries. With the rapid growth of spatial transcriptome (ST) techniques that preserve the spatial locations of cells in addition to profiling gene expression, there is a pressing demand for new mathematical and computational methods to deal with the unique challenges associated with ST data for CCC inference. This application will focus on addressing three major unaddressed challenges for CCC inference using ST data obtained from a diverse set of current experimental techniques. The first aim is to use scRNA-seq data to a) improve the coverage of genes that are associated with ligands or receptors not well measured in ST data through novel Optimal Transport methods, and b) impute spot-resolution data using physical models to estimate gene expression level for individual cells in the spot – critical information needed for CCC inference. The second aim is to develop a comprehensive CCC inference method accounting for various CCC regulators, co-factors, regulated genes, and potential external signals by incorporating prior knowledge and additional data. The third aim is to create a host of tools by using network analysis methods and neural graph network methods for pattern recognition, systematic comparisons, and classification of spatial CCC networks inferred from ST data. The study premise is based on our novel and extensive preliminary results in CCC inference. The proposed studies are significant because they will create the first comprehensive integrated tool that can impute ST data, infer CCC, and classify CCC networks in a systematic way, and success of the studies will establish a new cornerstone for ST data analysis, leading to novel spatial biological insights for tissues. The proposed studies are innovative because the proposed tools will have novel functionalities that use the ST data to derive crucial biological information which is currently impossible to obtain. They will also result in several novel mathematical and computational methods in the areas of multiscale modeling, optimal transport, and deep learning that will have broad applications in single-cell and spatial genomics data analysis and beyond.