SPOTlight

SPOTlight deconvolves spatial transcriptomics (ST) spots by integrating ST with single-cell RNA sequencing (scRNA-seq) to resolve cell-type and cell-state composition in complex tissues, including data from 10X Genomics Visium.

Key Features:

• Seeded Non-Negative Matrix Factorization (NMF) Regression: Uses seeded NMF regression initialized with cell-type marker genes and non-negative least squares (NNLS) optimization to estimate the composition of each ST spot.
• Integration with scRNA-seq Data: Integrates spatial transcriptomics with scRNA-seq reference data to improve resolution of cell types and states.
• High Prediction Accuracy: Demonstrates robust prediction accuracy, including with shallowly sequenced or small scRNA-seq reference datasets.
• Flexible Application Spectrum: Has been applied to map neuronal cell states and cortical/hippocampal architecture in mouse brain and to segment human pancreatic cancer tissue.
• Clinical Relevance: Localizes tumor-specific and clinically relevant immune cell states by training on external single-cell pancreatic tumor references.

Scientific Applications:

• Neuroscience: Mapping neuronal cell states and cortical layer or hippocampal architecture in the mouse brain using integrated ST and scRNA-seq data.
• Oncology: Segmenting tumor sections and distinguishing neoplastic versus normal cells in human pancreatic cancer studies.
• Immunology: Localizing immune cell states within tissue sections to study tumor immune microenvironments and clinically relevant immune phenotypes.

Methodology:

Seeded NMF regression initialized with cell-type marker genes and NNLS optimization applied to spatial transcriptomics data integrated with scRNA-seq references (applicable to 10X Genomics Visium and other ST platforms).