TISSUE: uncertainty-calibrated prediction of single-cell spatial transcriptomics improves downstream analyses.

Area of Science:

• Single-cell biology
• Computational biology
• Genomics

Background:

• Accurate spatial gene expression profiling at single-cell resolution is crucial but challenging.
• Existing spatial gene expression prediction methods lack robust uncertainty estimation, limiting their reliability.

Approach:

• Developed TISSUE (Transcript Imputation with Spatial Single-cell Uncertainty Estimation), a general framework for quantifying prediction uncertainty.
• Integrated uncertainty estimation into downstream analyses like differential gene expression, clustering, and machine learning.

Key Points:

• TISSUE provides well-calibrated prediction intervals across diverse datasets.
• Consistently improves differential gene expression analysis by reducing false discovery rates.
• Enhances clustering, visualization, and supervised learning model performance using predicted spatial transcriptomics.

Conclusions:

• TISSUE offers a flexible, uncertainty-aware approach for spatial transcriptomics data analysis.
• Enables more reliable identification of cell subtypes and regional classifiers, as demonstrated in mouse brain data.
• Facilitates robust inference from predicted gene expression profiles.