Evaluating cell type deconvolution in FFPE breast tissue: application to benign breast disease.

Area of Science:

• Biomedical research
• Genomics
• Computational biology

Background:

• Bulk RNA sequencing (RNA-seq) of formalin-fixed paraffin-embedded (FFPE) tissues is standard but averages gene expression, hindering biomarker discovery in heterogeneous samples.
• Accurately defining cell type composition is crucial for interpreting bulk RNA-seq data and identifying reliable biomarkers.

Purpose of the Study:

• To evaluate cell type deconvolution methods for bulk FFPE breast tissue samples.
• To identify robust methods that can overcome FFPE-induced artifacts and accurately estimate cell proportions.
• To introduce a new R package for facilitating cell type deconvolution analysis.

Main Methods:

• Construction of single-cell RNA-seq reference data for breast tissue.
• Simulation experiments to assess deconvolution method performance under FFPE conditions.
• Application of deconvolution methods to a cohort of 62 RNA-seq benign breast disease samples.
• Utilized digital pathology for cell type composition estimation.
• Developed and implemented the SCdeconR R package.

Main Results:

• Most deconvolution methods showed reduced performance with FFPE artifacts (RMSE 0.04–0.17).
• The deep learning-based method, Scaden, demonstrated superior robustness against FFPE artifacts.
• Pre-filtering reference data improved accuracy for most methods, reducing RMSE by up to 32%.
• SCdeconR package provides a platform for deconvolution assessment and analysis.

Conclusions:

• Cell type deconvolution from FFPE samples is challenging due to artifacts, but Scaden offers a promising solution.
• Reference data pre-filtering significantly enhances deconvolution accuracy.
• The SCdeconR package will aid researchers in analyzing cell composition from FFPE tissues.