scCDC: a computational method for gene-specific contamination detection and correction in single-cell and...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Droplet-based single-cell and single-nucleus RNA sequencing (scRNA-seq) are powerful tools for analyzing gene expression at the individual cell level.
• A significant challenge in scRNA-seq is the contamination from ambient RNA molecules, which can lead to inaccurate gene expression quantification.
• Current methods for ambient RNA removal often apply corrections globally, lacking specific evaluation for varying contamination levels and potentially affecting non-contaminated genes.

Purpose of the Study:

• To evaluate the efficacy of existing ambient RNA decontamination methods across different contamination levels.
• To develop a novel computational method for accurate and targeted removal of ambient RNA contamination in scRNA-seq data.
• To address the limitations of global correction methods by focusing on specific contamination-causing genes.

Main Methods:

• Comparative analysis of existing decontamination tools (DecontX, CellBender, SoupX, scAR) against simulated and real scRNA-seq datasets with varying contamination levels.
• Development of scCDC (single-cell computational decontamination), a novel algorithm designed to detect contamination-implicated genes and perform targeted expression correction.
• Validation of scCDC's performance against established methods, assessing its ability to accurately decontaminate highly contaminated genes while preserving the integrity of non-contaminated ones.

Main Results:

• Existing methods like DecontX and CellBender show under-correction for highly contaminated genes.
• SoupX and scAR demonstrate over-correction issues, particularly impacting lowly or non-contaminating genes.
• scCDC successfully identifies contamination-causing genes, including some cell-type markers, and selectively corrects their expression levels.
• scCDC significantly outperforms existing methods in accurately removing high-level contamination without introducing over-correction bias in other genes.

Conclusions:

• Ambient RNA contamination poses a critical challenge in scRNA-seq, necessitating precise correction strategies.
• scCDC represents a significant advancement, offering targeted decontamination that enhances the reliability of gene expression quantification.
• The development of scCDC provides researchers with a more accurate tool for analyzing scRNA-seq data, especially in the presence of significant ambient RNA contamination.