Multimodal joint deconvolution and integrative signature selection in proteomics.

Area of Science:

• Multi-omics data analysis
• Computational biology
• Proteomics and transcriptomics integration

Background:

• Cell-type-specific (cs) transcriptomic deconvolution often requires specific reference profiles.
• Existing methods face limitations in dissecting bulk proteome without dedicated proteomics reference panels.

Purpose of the Study:

• To present a novel method for dissecting bulk proteome using tissue-matched transcriptome and proteome.
• To identify proteins contributing to cellular heterogeneity shared between bulk transcriptome and proteome.
• To enable downstream analyses like cell-specific protein Quantitative Trait Loci (cspQTL) mapping.

Main Methods:

• Leveraging tissue-matched transcriptome and proteome data for deconvolution.
• Developing a multimodal approach without a proteomics reference panel.
• Benchmarking using CITE-seq pseudo-bulk data, simulations, and human brain/breast cancer tissues.

Main Results:

• Robust and accurate cell abundance quantification across diverse datasets.
• Successful identification of proteins contributing to shared cellular heterogeneity.
• Demonstrated utility for cspQTL mapping and multi-omics visualization.

Conclusions:

• The presented method offers an efficient way to dissect bulk proteome and analyze cell-specific protein signals.
• The MICSQTL tool facilitates advanced multi-omics analyses, including cspQTL mapping.
• This approach enhances the understanding of cellular heterogeneity in complex biological systems.