Harmony-R

Harmony-R integrates single-cell RNA sequencing (scRNA-seq) datasets by correcting batch effects and projecting cells into a shared embedding to group cells by biological cell type across technologies and experimental conditions.

Key Features:

• Batch correction and meta-analysis: Performs batch-effect correction across multiple scRNA-seq datasets to enable combined analyses and meta-analyses.
• Shared embedding space: Projects cells into a unified embedding where biological similarity, including cell type, is prioritized over dataset-specific variation.
• Scalability and efficiency: Scales to datasets containing up to ~10^6 cells and reduces computational resource requirements compared to some existing algorithms.
• Multifactor integration: Models multiple experimental and biological covariates simultaneously during integration to disentangle technical and biological variation.
• Multimodal integration: Supports integration across modalities, including combining scRNA-seq with spatial transcriptomics data.

Scientific Applications:

• Meta-analysis of scRNA-seq datasets: Enables combined analysis of data from multiple studies to integrate heterogeneous scRNA-seq datasets.
• Immunogenomics (PBMC integration): Used to integrate peripheral blood mononuclear cell (PBMC) datasets with significant experimental differences.
• Developmental biology (mouse embryogenesis): Applied to integrate mouse embryogenesis datasets for comparative analysis across developmental conditions.
• Tissue-specific studies (pancreatic islet cells): Applied to integrate pancreatic islet cell datasets across experiments.
• Spatial transcriptomics integration: Facilitates combination of scRNA-seq with spatial transcriptomics data.
• Disease modeling: Supports studies that require comprehensive transcriptional characterization of cell types across different disease conditions.

Methodology:

Performs batch-effect correction by projecting cells into a shared embedding space while simultaneously modeling multiple experimental and biological covariates.