Probabilistic cell/domain-type assignment of spatial transcriptomics data with SpatialAnno.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Cell type classification is crucial for analyzing single-cell RNA sequencing (scRNA-seq) and spatially resolved transcriptomics (SRT) data.
• Existing annotation methods primarily focus on scRNA-seq and neglect spatial information.

Purpose of the Study:

• To develop SpatialAnno, an efficient and accurate method for annotating spatial transcriptomics datasets.
• To effectively utilize non-marker genes and qualitative marker gene information without a reference dataset.

Main Methods:

• SpatialAnno employs a factor model to estimate low-dimensional embeddings for numerous non-marker genes.
• A Potts model is utilized to promote spatial smoothness among neighboring spots.
• The method was validated using simulated and real SRT datasets from various platforms (10x Visium, ST, Slide-seqV1/2, seqFISH).

Main Results:

• SpatialAnno demonstrates improved spatial annotation accuracy across diverse SRT datasets.
• The method shows robustness to irrelevant marker genes and marker gene misspecification.
• SpatialAnno is computationally scalable and platform-agnostic.

Conclusions:

• SpatialAnno provides an efficient and accurate solution for cell/domain type annotation in spatial transcriptomics.
• The method's ability to leverage non-marker genes and spatial context enhances biological insights.
• Estimated embeddings facilitate downstream analyses of cellular biological effects.