SpatialArtifacts: a computational framework for tissue artifact detection in spatial transcriptomics data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics data often contain technical artifacts like dry patches and uneven coverage.
• These artifacts, particularly at tissue borders, lead to low unique molecular identifier (UMI) counts and can be difficult to detect with current methods.

Purpose of the Study:

• To introduce SpatialArtifacts, a novel computational framework for identifying and classifying spatial artifacts in transcriptomics data.
• To provide a robust method for distinguishing true biological signals from technical noise in spatial datasets.

Main Methods:

• Employs median absolute deviation (MAD)-based outlier detection combined with mathematical morphology operations.
• Utilizes focal operations (fill, outline, star-pattern connectivity) to identify low-quality spots while preserving biological domains.
• Implements a hierarchical classification system to differentiate artifact types (edge/interior, large/small).

Main Results:

• Successfully identified and classified spatially contiguous tissue artifacts across diverse human tissues (hippocampus, prefrontal cortex, colorectal cancer).
• Demonstrated effectiveness on 10x Genomics Visium and VisiumHD platforms.
• The method effectively distinguishes artifacts from genuine biological regions.

Conclusions:

• SpatialArtifacts offers a reliable solution for addressing technical artifacts in spatial transcriptomics.
• The framework enables improved data quality, facilitating more accurate downstream analysis and interpretation.
• The SpatialArtifacts package is publicly available for use in the research community.