A comparative study of RNA-seq analysis strategies.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA sequencing (RNA-seq) is crucial for understanding gene expression.
• Accurate inference of expressed transcripts is essential for reliable RNA-seq analysis.
• Current methods for transcript set estimation include using curated annotations, genome-guided assembly, and de novo assembly.

Purpose of the Study:

• To systematically assess the performance of three principal RNA-seq transcript inference approaches.
• To evaluate the sensitivity, precision, and signal-to-noise ratios of different computational methods.
• To identify the most robust approach for transcript set estimation in RNA-seq data.

Main Methods:

• A simulation study was conducted to systematically compare three transcript inference approaches.
• Approaches evaluated: 1) using curated annotations, 2) genome-guided assembly, and 3) de novo assembly.
• Performance metrics included sensitivity, precision, and signal-to-noise ratios.

Main Results:

• Computational transcript set estimation exhibits severely limited sensitivity.
• Genome-guided and de novo assembly methods generate numerous artefacts with high expression estimates, absorbing significant biological signal.
• The curated annotation approach demonstrates good expression correlation, even with incomplete annotation databases.

Conclusions:

• Curated annotations provide a more reliable method for transcript set estimation in RNA-seq compared to computational assembly methods.
• Prioritizing sensitivity over precision in curated annotation sets is beneficial for robust expression analysis.
• Available software allows for simulation and comparison of transcript inference methods to aid further research.