Multimodal deep learning approaches for single-cell multi-omics data integration.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell multi-omics data integration is crucial for understanding complex cellular systems.
• Deep learning (DL) offers superior performance over classical methods for multi-omics integration.
• A systematic review of DL applications in single-cell multi-omics integration is lacking.

Approach:

• Conducted a comprehensive literature review on multimodal deep learning for single-cell multi-omics data integration.
• Summarized various data modalities in single-cell multi-omics.
• Categorized DL-based integration methods by data modality, DL architecture, fusion strategy, and tasks.

Key Points:

• Deep learning techniques are increasingly used for single-cell multi-omics integration.
• Methods are categorized based on data types, network architectures, and how data is combined.
• The review covers tasks like data imputation, cell type classification, and trajectory inference.

Conclusions:

• Deep learning models provide powerful tools for integrating diverse single-cell multi-omics data.
• This systematic approach aids in a deeper understanding of cellular heterogeneity and biological mechanisms.
• Future research can leverage these insights for advanced multi-omics data analysis.