scProca: A Cross-Attention-Enhanced Deep Generative Model for Single-Cell Transcriptomics and Proteomics Integration...

Area of Science:

• Computational Biology
• Genomics
• Proteomics

Background:

• Understanding complex diseases requires analyzing cellular interactions and protein expression.
• Large-scale sequencing aids disease subtyping, but integrating multi-omics data offers deeper cellular insights.
• Current models often use simplistic approaches for joint transcriptomics and proteomics analysis.

Purpose of the Study:

• To introduce scProca, a deep generative model for advanced integration of single-cell RNA sequencing (scRNA-seq) and co-profiling data.
• To address limitations in current models for handling heterogeneous biological data and inter-cellular relationships.

Main Methods:

• Developed scProca, a deep generative model utilizing cross-attention mechanisms.
• Incorporated inter-cellular relationships to handle heterogeneous inputs from RNA-seq and co-profiling datasets.
• Evaluated model performance on integration, imputation, robustness to sparsity, cross-species generalization, scalability, and batch compatibility.

Main Results:

• scProca achieves state-of-the-art performance in data integration and imputation.
• The model demonstrates robustness even with high protein sparsity.
• scProca generalizes across different species and tissues and scales effectively to large datasets.

Conclusions:

• scProca offers a flexible and powerful approach for integrating multi-omics data in complex biological systems.
• The model's ability to handle heterogeneous inputs and inter-cellular relationships advances single-cell analysis.
• scProca is well-suited for complex experimental settings, including those with batch effects.