SPAMMER: Spatial Analysis of Multi-omics Measurements in R
Author(s): Harkirat Kaur Sohi,Jason E. McDermott,Tong Zhang,Tujin Shi,Sara Jane Gosline
Affiliation(s): Pacific Northwest National Labs
Most omics-level technologies fail to properly characterize intra-tissue heterogeneity, as sample processing steps require homogenization that confounds spatial signatures. While there are existing tools for general multi-omics data processing, tools for spatially resolved omics data are limited. Further, such tools are almost non-existent for spatial proteomics data. To address this need, we introduce SPAMMER: Spatial Analysis of Multi-omics Measurements in R, a package that provides the data structure, functions, and workflow for end-to-end analysis of spatial data for different omics types. Though SPAMMER is applicable to spatial data of multiple omic types, we have focused specifically on spatial proteomics for the demonstration of the package. Spatial proteomics is a burgeoning field that measures over 6000 proteins in ~200 µm x 200 µm regions of a tissue sample while still retaining the spatial information of each region. We describe the overall functionality of SPAMMER as well as demonstrate its usefulness in describing spatial heterogeneity of human organ tissue using real-word examples of spatial proteomics data. Specifically, SPAMMER leverages existing spatial data processing and omics-specific tools such as the SingleCellExperiment object, and extends these to meet the needs of spatial omics data analysis. These functions include: 1) deriving signature features (example: proteins) for regions of interest within a spatial environment, 2) distance gradient analysis which can include identifying the correlation between physical distance from a location and a measurement of protein or gene expression or other functional terms from omics analysis, 3) representing spatial information of the regions of interest through adjacency graphs and 4) visualizing results from functional enrichment on the adjacency graphs and/or also onto a physical spatial map depending on the question of interest and network analysis. In addition to these novel tools, SPAMMER includes many of the standard procedures for analyzing omics data: potential solutions to dealing with missing data, data transformations and clustering techniques such as Principal Component Analysis (PCA), batch correction, pathway enrichment and differential abundance analysis to compare data from different spatial locations or different groups of samples/individuals. We will show how SPAMMER provides a framework in which spatial information is best utilized in conjunction with proteomics data to enable end-to-end analysis of spatially resolved omics data. As the technology grows, we hope to build upon this package to meet the needs of the community. For example, we hope to extend our analysis framework from two to three dimensions, enable automatic image detection of the sampled tissue to obtain spatial coordinates of the regions of interest, and enhanced clustering of the regions based on molecular measurements.