Biomarker Identification from RNA-Seq Data using a Robust Statistical Approach.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• RNA sequencing (RNA-seq) is crucial for transcriptomics, enabling biomarker identification through differentially expressed genes (DEGs).
• Existing DEG identification methods (e.g., edgeR, SAMSeq, voom-limma) struggle with high-dimensional RNA-seq data, often yielding high false positives and low accuracy, especially with outliers.

Purpose of the Study:

• To introduce a robust t-statistic method for identifying DEGs from RNA-seq data.
• To overcome limitations of existing methods, particularly in the presence of data outliers.
• To validate the robust t-statistic method using simulated and real RNA-seq datasets.

Main Methods:

• Development and application of a robust t-statistic method for DEG analysis.
• Utilized simulated and real RNA-seq datasets, including a dataset comparing HIV viremic vs. aviremic states.
• Performance evaluation included sensitivity, specificity, MER, FDR, AUC, ACC, PPV, and NPV, particularly with 20% outliers. Protein-protein interaction (PPI) analysis was performed using the STRING database.

Main Results:

• The robust t-statistic method demonstrated improved performance metrics, achieving 74.5% AUC and 78.4% ACC with 20% outliers.
• Identified 409 DEGs (p<0.05) in the HIV dataset, with 28 up-regulated and 381 down-regulated genes (log2 fold change threshold 1.5).
• Discovered 11 up-regulated genes strongly associated with HIV-1/AIDS and identified 21 genes with significant interactions via PPI analysis.

Conclusions:

• The robust t-statistic model effectively identifies DEGs and potential biomarkers from RNA-seq data, outperforming traditional methods in the presence of outliers.
• This method enhances the reliability of transcriptomic data analysis for biomarker discovery, particularly in complex biological conditions like HIV/AIDS.