Artifacts in spatial transcriptomics data: their detection, importance, prevalence, and prevention

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics technologies enable gene expression analysis within tissue context.
• Data artifacts, such as border effects and batch malfunctions, can compromise the accuracy of spatial transcriptomics findings.
• Existing quality control methods may not adequately address these specific artifact types across different platforms.

Purpose of the Study:

• To develop and validate an automated, cross-platform method for detecting and removing data artifacts in spatial transcriptomics.
• To address common artifacts including border effects, tissue edge effects, and location batch malfunctions.
• To improve the reliability and reproducibility of spatial transcriptomics data analysis.

Main Methods:

• Development of Border, Location, and edge Artifact DEtection (BLADE), a suite of automated statistical methods.
• Application of BLADE to diverse spatial transcriptomics platforms, including Visium and CosMx.
• Evaluation of BLADE using a library of 37 10x Visium samples from human and mouse liver and adipose tissues.

Main Results:

• BLADE successfully detected and enabled the removal of border effects, tissue edge effects, and location batch malfunctions.
• Artifacts were found to be prevalent and significantly impactful in the evaluated spatial transcriptomics samples.
• The study demonstrated the critical need for robust artifact detection in spatial transcriptomics quality control.

Conclusions:

• BLADE provides a novel, automated solution for identifying and mitigating common artifacts in spatial transcriptomics data.
• The developed software is cross-platform compatible and publicly available, facilitating broader adoption in the research community.
• Implementing artifact detection methods like BLADE is essential for ensuring the integrity of spatial transcriptomics findings.