scJVAE: A novel method for integrative analysis of multimodal single-cell data

Area of Science:

• Single-cell omics
• Computational biology
• Genomics

Background:

• Multimodal single-cell omics technology enables deeper insights into cellular characteristics by profiling multiple data types from the same cell.
• Learning joint representations of multimodal single-cell data is hindered by technical variations known as batch effects.
• Existing methods struggle to effectively integrate and correct for batch effects in complex single-cell datasets.

Purpose of the Study:

• To introduce scJVAE (single-cell Joint Variational AutoEncoder), a novel computational method for batch effect removal and joint representation learning.
• To integrate and learn a joint embedding from paired single-cell RNA sequencing (scRNA-seq) and single-cell Assay for Transposase-Accessible Chromatin sequencing (scATAC-seq) data.
• To evaluate the performance of scJVAE in removing batch effects and facilitating downstream analyses.

Main Methods:

• Development of the scJVAE model, a Joint Variational AutoEncoder specifically designed for multimodal single-cell data.
• Integration of paired scRNA-seq (gene expression) and scATAC-seq (open chromatin) data modalities within the scJVAE framework.
• Evaluation of scJVAE's batch effect removal capabilities across diverse datasets and assessment of its performance in downstream tasks like clustering.

Main Results:

• scJVAE effectively removes batch effects from multimodal single-cell datasets, including paired gene expression and open chromatin data.
• The method successfully integrates scRNA-seq and scATAC-seq data, learning a robust joint representation.
• scJVAE demonstrates superior performance compared to existing state-of-the-art methods in batch effect correction and data integration.

Conclusions:

• scJVAE is a robust and scalable method for addressing batch effects in multimodal single-cell omics data.
• The developed approach enhances downstream analyses such as dimensionality reduction and cell-type clustering.
• scJVAE offers a significant advancement for researchers working with integrated single-cell datasets, improving data quality and analytical outcomes.