Harmonizing heterogeneous single-cell gene expression data with individual-level covariate information.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data integration is crucial for identifying shared and unique cellular features across samples.
• Existing integration methods often struggle with technical variations, biological differences, and accounting for individual-level covariates (e.g., age, disease status).
• Many current approaches are limited to discrete variables, hindering comprehensive analysis.

Purpose of the Study:

• To develop a robust method for harmonizing scRNA-seq datasets that accounts for both continuous and discrete individual-level covariates.
• To improve the accuracy and biological relevance of integrated scRNA-seq analyses.
• To provide a flexible tool for researchers analyzing diverse scRNA-seq data.

Main Methods:

• Proposed scINSIGHT2, a generalized linear latent variable model.
• The model accommodates continuous covariates (e.g., age) and discrete factors (e.g., disease conditions).
• Validated through simulation studies and real-world scRNA-seq data applications.

Main Results:

• scINSIGHT2 demonstrated accurate harmonization of scRNA-seq datasets from single and multiple sources.
• The method effectively captures meaningful biological insights.
• Results show the utility of scINSIGHT2 in handling individual-level variations within scRNA-seq data.

Conclusions:

• scINSIGHT2 offers a powerful and flexible approach for scRNA-seq data integration.
• The method successfully addresses limitations of existing tools by incorporating diverse covariates.
• scINSIGHT2 enhances the ability to derive biological insights from complex scRNA-seq datasets.