A sequential Monte Carlo approach to gene expression deconvolution

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput gene expression data from heterogeneous samples pose analytical challenges.
• Manual cell separation methods are time-consuming and prone to contamination.
• Accurate deconvolution of mixed cell signals is crucial for biological insights.

Purpose of the Study:

• To develop a robust computational method for analyzing gene expression data from heterogeneous biological samples.
• To accurately estimate cell type proportions and cell-type specific gene expression.
• To improve the identification of differentially expressed genes in complex tissues.

Main Methods:

• A Bayesian framework was employed to model heterogeneous gene expression data.
• A novel sequential Monte Carlo (SMC) sampler was developed for parameter estimation.
• The method approximates posterior distributions using weighted samples in a high-dimensional space.

Main Results:

• The proposed SMC algorithm demonstrated superior performance on simulated and real datasets.
• Improved accuracy was achieved in estimating cell type proportions and cell-type specific expressions.
• More accurate identification of differentially expressed genes compared to Dsection and NMF.

Conclusions:

• The SMC-based Bayesian framework offers an efficient and accurate approach for deconstructing heterogeneous gene expression data.
• This method enhances the analysis of complex biological samples, overcoming limitations of traditional techniques.
• The algorithm provides a valuable tool for genomic research, with MATLAB implementation available.