Gene count normalization in single-cell imaging-based spatially resolved transcriptomics

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Imaging-based spatially resolved transcriptomics (im-SRT) allows high-throughput gene profiling in fixed tissues.
• Normalization is essential in im-SRT to correct for technical variations and ensure accurate biological signal interpretation.

Approach:

• Investigated the impact of various gene count normalization methods on im-SRT data.
• Utilized simulated gene panels overrepresenting specific tissue regions or cell types.
• Compared normalization effects across different gene panels and their representativeness.

Key Points:

• Gene count-based normalization methods can introduce region- or cell type-specific biases in normalized gene expression magnitudes.
• These biases may lead to unreliable downstream analyses, including differential gene expression and spatially variable gene analysis, causing false positives/negatives.
• Normalization methods not relying on detected gene counts (e.g., cell volume/area) avoid these observed effects.

Conclusions:

• Recommends using non-gene count-based normalization approaches for im-SRT data when possible.
• Emphasizes evaluating gene panel representativeness before applying gene count-based normalization.
• Highlights that normalization method and gene panel choice significantly influence the biological interpretation of im-SRT data.