scMODAL: a general deep learning framework for comprehensive single-cell multi-omics data alignment with feature...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell technologies offer deep insights into cellular states via transcriptomic, epigenomic, and proteomic profiling.
• Integrating multi-omics data is crucial for understanding complex cell functions and disease mechanisms.
• Existing computational methods face challenges in aligning diverse single-cell omics datasets due to technical limitations and varied feature correlations.

Purpose of the Study:

• To introduce scMODAL, a novel deep learning framework for single-cell multi-omics data integration.
• To address the challenges of aligning datasets with limited known feature correlations.
• To provide a robust computational solution for advancing single-cell multi-omics research.

Main Methods:

• scMODAL utilizes a deep learning framework incorporating neural networks and generative adversarial networks.
• The method employs feature links for aligning cell embeddings across different omics modalities.
• It is designed to handle datasets with limited positively correlated features.

Main Results:

• scMODAL effectively removes unwanted variation while preserving biological information in integrated datasets.
• The framework accurately identifies cell subpopulations across diverse single-cell multi-omics datasets.
• Experiments demonstrate the framework's capability in data alignment and preserving feature topology.

Conclusions:

• scMODAL offers a powerful approach for integrating single-cell multi-omics data, even with limited feature correlations.
• The framework facilitates downstream analyses, including feature imputation and relationship inference.
• scMODAL represents a significant advancement for computational analysis in single-cell multi-omics research.