Disentangling Heritable Variation from Microenvironmental Regulation in Spatial Lineage Tracing Data with SOUR

Intratumoral heterogeneity is a major barrier to cancer therapy, yet the processes that generate it remain only partially understood. Recently developed single cell technologies that simultaneously profile gene expression, lineage information, and spatial positions offer new opportunities to study tumor evolution. A central question for such data is how much of the expression variation in a gene is heritable versus driven by the local microenvironment. Currently, this question is answered with marginal statistics---for example, the autocorrelation of expression with lineage distance and, separately, with spatial distance---which are confounded because closely related cells also tend to be spatially proximate. Here, we present a new model that simultaneously regresses gene expression on lineage and spatial information, enabling estimation of the unique contribution of each component while controlling for the other.