Skip to content

Glossary

Key terms used throughout this site, listed alphabetically.

AnnData A data structure (Python, part of the scverse ecosystem) for storing annotated matrices. The standard container for single-cell and spatial omics data in Python workflows. Stores expression matrices, cell/gene metadata, embeddings, and spatial coordinates in a single .h5ad file.

Barcode A short nucleotide sequence used to tag molecules or locations. In spatial transcriptomics, barcodes are attached to specific spatial positions on a capture array (Visium, Slide-seq) to link transcripts to their tissue coordinates.

Cell segmentation The process of assigning pixels or transcripts to individual cells in imaging-based spatial data. A critical preprocessing step that determines cell boundaries and directly affects all downstream analyses. See Segmentation Benchmarks.

Cell2location A Bayesian deconvolution method that estimates cell-type abundance at each spatial location using a reference single-cell RNA-seq dataset. Consistently ranked as the most accurate deconvolution method in benchmarks. See Deconvolution Benchmarks.

CCC (Cell-Cell Communication) Analysis of signaling interactions between cells, typically inferred from ligand-receptor co-expression patterns. Spatial CCC methods (COMMOT, SpatialDM, MISTy) additionally require that interacting cells be spatially proximate.

CODEX Co-Detection by Indexing. A cyclic immunofluorescence technology for spatial proteomics, now commercialized as PhenoCycler by Akoya Biosciences. Measures 40--100 proteins at single-cell resolution by iterative antibody staining and imaging.

CosMx CosMx Spatial Molecular Imager. A single-molecule imaging platform developed by NanoString (now Bruker) for spatial transcriptomics at subcellular resolution. Measures up to 6,000 genes using fluorescent probe hybridization.

DDA / DIA Data-Dependent Acquisition / Data-Independent Acquisition. Mass spectrometry acquisition strategies used in spatial proteomics (MALDI) and metabolomics. DIA provides more comprehensive coverage; DDA provides higher sensitivity for targeted analytes.

Deconvolution The computational process of estimating cell-type proportions within a spatial spot or region that contains multiple cells. Necessary for technologies like Visium where each spot captures RNA from 5--20 cells. See Deconvolution Benchmarks.

DLPFC Dorsolateral prefrontal cortex. A brain region used as the most common benchmark dataset for spatial domain detection methods (Maynard et al. 2021). Contains well-defined cortical layers (L1--L6 plus white matter). See Benchmark Synthesis for caveats about over-reliance on this dataset.

FISH Fluorescence In Situ Hybridization. A molecular biology technique that uses fluorescent probes to detect specific nucleic acid sequences in tissue. The basis for spatial transcriptomics technologies like MERFISH, seqFISH, and osmFISH.

Foundation model A large-scale machine learning model pre-trained on massive datasets that can be fine-tuned for specific tasks. In spatial omics, foundation models (e.g., scGPT, Geneformer, scBERT) trained on millions of cells are being adapted for spatial tasks like cell-type annotation and imputation.

GeoMx GeoMx Digital Spatial Profiler. A region-of-interest (ROI) spatial profiling platform developed by NanoString (now Bruker). Measures gene expression or protein levels in user-selected tissue regions rather than at single-cell resolution.

GNN (Graph Neural Network) A class of deep learning models that operate on graph-structured data. In spatial omics, GNNs construct graphs where cells/spots are nodes and spatial proximity defines edges, enabling methods (GraphST, STAGATE) to learn spatially-aware representations.

H&E Hematoxylin and Eosin. A standard histological stain that highlights tissue morphology. Many spatial transcriptomics platforms (Visium, Visium HD) capture H&E images alongside molecular data, enabling integration of morphological and molecular information.

IMC (Imaging Mass Cytometry) A spatial proteomics technology that uses metal-tagged antibodies and mass spectrometry to image 40+ proteins simultaneously at ~1 micron resolution. Commercialized by Standard BioTools (formerly Fluidigm).

In situ Latin for "in its place." Refers to measurements made within the original tissue context, preserving spatial relationships. Spatial omics technologies measure molecules in situ rather than after tissue dissociation.

ISS (In Situ Sequencing) A spatial transcriptomics approach that sequences RNA molecules directly within tissue using padlock probes and rolling circle amplification. Provides single-molecule resolution but with lower gene throughput than MERFISH.

Ligand-receptor A pair of molecules (one secreted or surface-bound, one on the receiving cell) that mediate cell-cell signaling. Ligand-receptor databases (CellPhoneDB, CellChat) form the basis of most cell-cell communication analysis methods.

MALDI Matrix-Assisted Laser Desorption/Ionization. A mass spectrometry technique used for spatial metabolomics and proteomics. Provides untargeted molecular profiling at 5--50 micron resolution without requiring antibodies or probes.

MERFISH Multiplexed Error-Robust Fluorescence In Situ Hybridization. A single-molecule imaging technology that measures hundreds to thousands of genes at subcellular resolution. Commercialized as the MERSCOPE platform by Vizgen (now 10x Genomics).

Niche A local cellular neighborhood in tissue, defined by the composition and spatial arrangement of cell types. Niche detection identifies recurring spatial patterns of cell-type co-localization that may reflect functional tissue units.

Optimal transport A mathematical framework for finding the most efficient mapping between two distributions. Used in spatial omics for mapping single-cell reference data to spatial locations (Tangram, novoSpaRc) and modeling intercellular communication (COMMOT).

PhenoCycler The commercial name for the CODEX technology, sold by Akoya Biosciences. A cyclic immunofluorescence platform for high-plex spatial proteomics.

Resolution The spatial granularity of a measurement. In spatial omics, resolution ranges from ~55 microns (original Visium spots, containing 5--20 cells) to ~100 nanometers (MERFISH, Xenium, where individual transcripts are localized). Higher resolution provides more spatial detail but generates larger datasets.

scRNA-seq Single-cell RNA sequencing. The dissociated single-cell transcriptomics approach that measures gene expression in individual cells without preserving spatial information. Frequently used as a reference for spatial deconvolution and cell-type annotation.

Spatial domain A contiguous region of tissue with a coherent gene expression program. Spatial domain detection (also called spatial clustering) identifies these regions by combining transcriptomic similarity with spatial contiguity. See Clustering Benchmarks.

Spot A discrete capture location on a spatial array. In Visium, spots are 55-micron circles arranged in a hexagonal grid, each capturing RNA from 5--20 cells. In Visium HD, bins are 2 microns. The term "spot" is sometimes used loosely for any spatial measurement unit.

SVG (Spatially Variable Gene) A gene whose expression pattern shows spatial structure beyond random expectation. SVGs mark tissue domains, gradients, or localized cell states. Detected using spatial autocorrelation methods (nnSVG, SPARK-X, SpatialDE). See SVG Benchmarks.

Visium A sequencing-based spatial transcriptomics platform by 10x Genomics. Captures mRNA at ~3,500 spots (55-micron diameter) across a tissue section with genome-wide coverage. The most widely used spatial transcriptomics technology and the most common benchmark platform.

Visium HD The high-resolution successor to Visium, capturing at 2-micron bins. Provides near single-cell resolution with genome-wide coverage but generates much larger datasets. See The Technology-Analysis Gap for analysis challenges.

Xenium An in situ imaging platform by 10x Genomics that measures targeted gene panels (up to ~5,000 genes) at subcellular resolution. Provides single-cell and subcellular transcript localization with high sensitivity. Competes with CosMx and MERSCOPE in the imaging-based spatial transcriptomics market.