Bayesian log-normal deconvolution for enhanced in silico microdissection of bulk gene expression data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Deconvolving bulk gene expression is crucial for understanding complex tissues, including the tumor microenvironment.
• Accurate deconvolution requires robust statistical models and prior single-cell RNA sequencing data due to gene expression variability.

Purpose of the Study:

• Introduce BLADE (Bayesian Log-normAl Deconvolution), a unified Bayesian framework for estimating cellular composition and cell-type-specific gene expression profiles.
• Develop a method capable of handling a large number of cell types (>20) efficiently.

Main Methods:

• Utilizes a unified Bayesian framework with efficient variational inference.
• Designed to integrate prior knowledge from single-cell RNA sequencing data.

Main Results:

• BLADE demonstrated enhanced robustness against gene expression variability compared to conventional methods.
• Achieved better completeness in reconstructing gene expression profiles for individual cell types.
• Successfully handled datasets with over 20 cell types due to efficient variational inference.

Conclusions:

• BLADE is a powerful tool for unraveling heterogeneous cellular activity in complex biological systems using bulk gene expression data.
• The method improves the reconstruction of cell-type-specific gene expression profiles, crucial for systems biology and disease research.