PROJECT SUMMARY Advances in single cell assays have enabled the genome-wide measurement of DNA sequence, RNA expression, chromatin accessibility and protein abundance for tens-of-thousands of cells isolated from a single tissue sample. Methods that capture or infer spatial or temporal information provide additional contextual information to create a detailed, cell-level picture of gene activity and function. These meth- ods give researchers a powerful tool for identifying the cell types in the analyzed tissue, the phenotype of those cells and the network of cell-cell interactions that control tissue structure and function. Although these techniques have revolutionized the study of complex tissues, the significant sparsity and noise of single cell measurements means that statistical analysis is typically performed at the level of large groups or clusters of cells. Although a cluster-based analysis can mitigate sparsity and noise, the re- sults reflect the state of the average cell in the cluster, which may be quite dissimilar from many cells in heterogeneous populations. To fully realize the potential of single cell profiling, bioinformatics meth- ods are needed that can accurately characterize individual cells rather than cell groups. One promising approach for generating cell-level estimates is gene set testing or pathway analysis, which can more effectively capture the state of individual cells by combining the measurements for all genes in a biolog- ical pathway. Unfortunately, statistical and biological differences between single cell and bulk genomic data make it challenging to use existing gene set testing methods, that were developed for bulk tissue, on single cell data. To address this limitation, we will create innovative algorithms for cell-level gene set testing and will use these techniques to support the estimation of cell type, phenotype and interaction potential. The translational application of these methods to study immune cell signaling within the tu- mor microenvironment will help validate our approach and provide important insights into the immune response to cancer.

PROJECT NARRATIVE Recent advances in single cell assays have enabled the genome-wide profiling of tens-of-thousands of individual cells from a single tissue sample, however, the significant sparsity and noise of single cell measurements means that analysis of this data is typically performed at the level of large groups or clusters of cells. Unfortunately, cluster-based results reflect the state of the average cell in the cluster, which may be quite dissimilar from many cells in heterogeneous populations. To address this limitation, we will create innovative gene set-based methods that estimate cell type, phenotype and interaction potential for individual cells rather than cell groups.