Integration of Bulk RNA-seq Pipeline Metrics for Assessing Low-Quality Samples.

Area of Science:

• Biomedical research
• Genomics
• Bioinformatics

Background:

• RNA sequencing (RNA-seq) is a vital tool in biomedical research.
• Standardized quality control (QC) guidelines are needed for RNA-seq data.
• Identifying informative technical metrics for low-quality sample detection remains a challenge.

Purpose of the Study:

• To develop and validate a software tool for comprehensive RNA-seq quality control.
• To identify the most informative QC metrics for assessing sample quality.
• To provide practical guidance for improving RNA-seq methodological rigor.

Main Methods:

• Development of the Quality Control Diagnostic Renderer (QC-DR) software.
• Application of QC-DR to a large clinical RNA-seq dataset (SCRIPT).
• Utilized machine learning models to predict sample quality based on QC metrics.

Main Results:

• QC-DR simultaneously visualizes multiple QC metrics and flags aberrant samples.
• Pipeline QC metrics like uniquely aligned reads, rRNA reads, detected genes, and AUC-GBC were highly correlated with sample quality.
• Experimental QC metrics showed no significant correlation with sample quality.
• Machine learning models trained on the SCRIPT dataset accurately predicted sample quality on an independent dataset.

Conclusions:

• Individual QC metrics have limited predictive value for RNA-seq sample quality.
• Integrating multiple QC metrics with defined thresholds is recommended.
• The developed QC-DR software offers practical guidance and improves the methodological rigor of RNA-seq studies.