xCell: Digitally portraying the tissue cellular heterogeneity landscape

Tissues are a complex milieu consisting of numerous cell types. For example, understanding the cellular heterogeneity the tumor microenvironment is an emerging field of research. Numerous methods have been published in recent years for the enumeration of cell subsets from tissue expression profiles. However, the available methods suffer from three major problems: inferring cell subset based on gene sets learned and verified from limited sources; displaying only partial portrayal of the full cellular heterogeneity; and insufficient validation in mixed tissues. To address these issues we developed xCell, a novel gene-signature based method for inferring 64 immune and stroma cell types. We first curated and harmonized 1,822 transcriptomic profiles of pure human cell types from various sources, employed a curve fitting approach for linear comparison of cell types, and introduced a novel spillover compensation technique for separating between closely related cell types. We test the ability of our model learned from pure cell types to infer enrichments of cell types in mixed tissues, using both comprehensive in silico analyses, and by comparison to cytometry immunophenotyping to show that our scores outperform previously published methods. Finally, we explore the cell type enrichments in tumor samples and show that the cellular heterogeneity of the tumor microenvironment uniquely characterizes different cancer types. We provide our method for inferring cell type abundances as a public resource to allow researchers to portray the cellular heterogeneity landscape of tissue expression profiles: http://xCell.ucsf.edu/.