TY - JOUR T1 - SpatialDE - Identification of spatially variable genes JF - bioRxiv DO - 10.1101/143321 SP - 143321 AU - Valentine Svensson AU - Sarah A Teichmann AU - Oliver Stegle Y1 - 2017/01/01 UR - http://biorxiv.org/content/early/2017/05/28/143321.abstract N2 - Technological advances have enabled low-input RNA-sequencing, paving the way for assaying transcriptome variation in spatial contexts, including in tissue systems. While the generation of spatially resolved transcriptome maps is increasingly feasible, computational methods for analysing the resulting data are not established. Existing analysis strategies either ignore the spatial component of gene expression variation, or require discretization.To address this, we have developed SpatialDE, a computational framework for identifying and characterizing spatially variable genes. Our method generalizes variable gene selection, as used in population- and single-cell studies, to spatial expression profiles. We apply SpatialDE to Spatial Transcriptomics and to data from single cells expression profiles using multiplexed In Situ Hybridisation (SeqFISH and MERFISH), demonstrating its general use. SpatialDE identifies genes with expression patterns that are associated with histology in breast cancer tissue, several of which have known disease implications and are not detected by variable gene selection. Additionally, our model can be used to classify genes with distinct spatial patterns, including periodic expression profiles, linear trends and general spatial variation ER -