omicsGMF: a multi-tool for dimensionality reduction, batch correction and imputation in bulk- and single-cell...

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Mass spectrometry (MS) enables large-scale proteomics and single-cell proteome profiling.
• Current MS data analysis faces challenges with batch effects and missing values.
• Existing workflows for dimensionality reduction are often sequential and lack scalability.

Purpose of the Study:

• To develop a fast, scalable, and interpretable matrix factorization method for proteomics data.
• To address limitations in dimensionality reduction, batch correction, and imputation for MS-based proteomics.
• To enhance the analysis of both bulk and single-cell proteomics datasets.

Main Methods:

• Introduced omicsGMF, a unified matrix factorization framework.
• Integrated imputation, batch correction, and principal component analysis into a single model.
• Applied omicsGMF to bulk and single-cell proteomics data.

Main Results:

• OmicsGMF provides robust imputation of missing values, surpassing existing tools.
• The integrated approach significantly enhances data processing and dimensionality reduction.
• Demonstrated increased statistical power for detecting differentially abundant proteins.

Conclusions:

• OmicsGMF offers a highly scalable solution for dimensionality reduction in proteomics.
• The method improves critical steps in proteomics data analysis, including imputation and batch correction.
• OmicsGMF facilitates more powerful downstream analyses of proteomics data.