Bulk brain tissue cell-type deconvolution with bias correction for single-nuclei RNA sequencing data using DeTREM.

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Accurate cell-type abundance quantification from bulk tissue RNA-sequencing is crucial for understanding complex biological systems.
• Current deconvolution methods like MuSiC often rely on single-cell RNA-sequencing (scRNA-seq) data for cell-type signatures.
• Challenges arise when using single-nuclei RNA-sequencing (snRNA-seq) data, particularly for tissues like the human brain, due to technological sequencing differences.

Purpose of the Study:

• To introduce DeTREM, a modified MuSiC algorithm designed to improve cell-type deconvolution accuracy.
• To address and compensate for sequencing discrepancies between snRNA-seq reference data and bulk RNA-seq datasets.
• To enhance the prediction of cell-type fractions in bulk tissue samples.

Main Methods:

• Development of DeTREM, an algorithm that modifies MuSiC to account for RNA-sequencing technology differences.
• Validation using simulated bulk RNA-sequencing datasets with varying cell-type compositions.
• Testing on real human brain bulk RNA-sequencing data.

Main Results:

• DeTREM demonstrates superior accuracy compared to the original MuSiC algorithm in both simulated and real human brain datasets.
• Comparative analysis shows DeTREM outperforms other leading deconvolution methods, SCDC and CIBERSORTx, on human brain data.
• While SCDC and CIBERSORTx perform well on simulated data, DeTREM shows greater robustness and accuracy with snRNA-seq derived signatures.

Conclusions:

• DeTREM significantly improves the accuracy of cell-type deconvolution when utilizing snRNA-seq data.
• The method provides reliable cell-type abundance estimates even when scRNA-seq data is unavailable.
• DeTREM facilitates better characterization of cell-type specific effects and abundance variations in brain tissue research.