Modular RNA-Sequencing Analytics for Exploratory Biomarker Discovery Using Public Data

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Publicly available RNA sequencing (RNA-seq) data are valuable for biomarker discovery but suffer from cross-study heterogeneity.
• Analyzing disparate RNA-seq datasets requires robust methods to standardize data and extract meaningful biological insights.

Purpose of the Study:

• To develop and demonstrate a modular analytics pipeline for unifying heterogeneous RNA-seq datasets.
• To enable robust biomarker discovery by integrating public data using open-source tools and machine learning.
• To showcase the pipeline's adaptability across different disease contexts and data types.

Main Methods:

• Developed a modular pipeline incorporating quality control, differential expression, and pathway analysis.
• Integrated competitive machine learning techniques to unify disparate RNA-seq datasets.
• Applied the pipeline to COVID-19 severity, multi-cohort sepsis, and atherosclerosis (bulk and single-cell) datasets.

Main Results:

• Successfully identified differentially expressed gene signatures for COVID-19 severity.
• Generated concise biomarker panels for sepsis using integrated differential expression and machine learning.
• Examined tissue and cell-type specificity of N-acyl-phosphatidylethanolamine phospholipase D in atherosclerosis.

Conclusions:

• An adaptable, modular pipeline using open-source tools can effectively repurpose public RNA-seq data.
• The pipeline reduces noise, generates novel hypotheses, and reveals biological insights for biomarker discovery.
• This approach establishes a foundation for future research leveraging public data.