A graph-based algorithm for RNA-seq data normalization.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• RNA sequencing (RNA-seq) is vital for biological system characterization.
• RNA-seq data normalization is challenging due to inherent circularity and reliance on prior assumptions.
• Existing methods fail with abundant and heterogeneous RNA-seq data.

Purpose of the Study:

• To develop a novel RNA-seq normalization procedure.
• To overcome the circularity problem without assuming non-differential transcripts.
• To provide accurate normalization for complex and heterogeneous RNA-seq datasets.

Main Methods:

• Developed a graph-based algorithm using intrinsic transcript correlations.
• Identified densely connected vertices within the transcript correlation graph as references.
• Validated the method on synthesized and ENCODE project datasets.

Main Results:

• The algorithm successfully recovered reference transcripts with high precision on synthesized data.
• Achieved high-quality normalization, outperforming assumption-based methods.
• Demonstrated good performance and reasonable runtime on a realistic ENCODE dataset.

Conclusions:

• The proposed method effectively addresses the RNA-seq normalization circularity problem.
• This approach offers a robust alternative for analyzing large, heterogeneous RNA-seq datasets.
• Potential to significantly advance RNA-seq data analysis and interpretation.