scTIE: data integration and inference of gene regulation using single-cell temporal multimodal data

Area of Science:

• Computational Biology
• Genomics
• Developmental Biology

Background:

• Single-cell technologies enable detailed gene regulatory analysis.
• Integrating multimodal single-cell data (scRNA-seq, scATAC-seq) is crucial but challenging for cell identification and regulatory inference.
• Existing methods often treat data integration and regulatory relationship extraction as separate problems.

Approach:

• Introduced scTIE, a unified computational method for temporal multimodal data integration and regulatory relationship inference.
• Employs an autoencoder with iterative optimal transport to embed cells from multiple time points into a shared latent space.
• Extracts interpretable information to predict cell trajectories and regulatory dynamics.

Key Points:

• scTIE effectively integrates temporal multimodal single-cell data, outperforming existing methods in preserving biological signals, especially with batch effects and noise.
• Demonstrated superior data integration and biological signal preservation on synthetic and real-world datasets.
• Identified key regulatory elements predictive of cell transition probabilities in differentiating mouse embryonic stem cells.

Conclusions:

• scTIE provides a unified framework for analyzing temporal multimodal single-cell data, advancing the understanding of gene regulatory networks.
• Offers new potential for dissecting the regulatory landscape driving complex biological processes like cell differentiation.
• Enables more accurate cell type identification and prediction of cellular state transitions.