A dual approach to evaluate the performance of RNA-Seq data analysis pipelines with weak signals

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• RNA-Seq (Ribonucleic acid sequencing) is a powerful tool for gene expression analysis.
• Selecting appropriate bioinformatics pipelines is critical for accurate RNA-Seq data interpretation.
• Challenges arise in identifying optimal pipelines for complex datasets, especially those with weak biological signals.

Purpose of the Study:

• To evaluate and rank 90 bioinformatics pipelines for RNA-Seq data analysis.
• To identify reliable pipelines capable of detecting weak biological signals from environmental exposures.
• To provide practical recommendations for selecting robust RNA-Seq analysis workflows.

Main Methods:

• A dual strategy was employed, combining RNA-Seq and qRT-PCR (quantitative Reverse Transcription Polymerase Chain Reaction) expression data correlation.
• Machine learning classifiers were utilized to rank pipelines based on their ability to distinguish between exposure groups.
• Performance evaluation focused on datasets from Rats, Zebrafish, and Mice exposed to metallic particles, low-dose radiation, or medical treatments.

Main Results:

• Pipeline selection significantly impacts RNA-Seq study efficiency and biological insight generation.
• The developed ranking methods are effective even with lower sequencing depth and noisy data.
• StringTie emerged as a recommended counting method for handling weak RNA-Seq signals.

Conclusions:

• Bioinformatics pipeline choice is critical for RNA-Seq studies, particularly those involving weak signals.
• The classifier-based ranking approach is valuable for identifying optimal pipelines across various research contexts.
• Prioritizing specific methods like StringTie for counting can enhance the reliability of RNA-Seq analysis for weak signals.