Alternative splicing detection workflow needs a careful combination of sample prep and bioinformatics analysis

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• RNA sequencing (RNA-Seq) is crucial for biomarker discovery and disease characterization.
• Library Sample Preparation (LSP) and Bioinformatics Analysis (BA) are critical steps influencing RNA-Seq outcomes.
• This study evaluates the combined impact of LSP methods and BA tools on splice variant detection.

Purpose of the Study:

• To assess how different Library Sample Preparation (LSP) methods affect splice variant detection in RNA-Seq.
• To compare the performance of various Bioinformatics Analysis (BA) tools for detecting alternative splicing events.
• To determine the influence of input RNA quantity and sequencing depth on the accuracy of splice variant identification.

Main Methods:

• Comparison of multiple LSP protocols including TruSeq (unstranded/stranded), ScriptSeq, and NuGEN.
• Utilized a benchmark dataset with spiked-in synthetic and real RNA-Seq reads from standard and low-input libraries.
• Evaluated splice variant detection using quantification tools (Cuffdiff2, RSEM-EBSeq) and exon-level analysis (DEXSeq).

Main Results:

• All tested LSPs identified a common set of splice variants, but each also detected unique low-coverage transcripts.
• Low-input RNA protocols (NuGEN v2) showed a pronounced effect on unique transcript detection.
• Exon-level analysis (DEXSeq) slightly outperformed splice variant quantification (Cuffdiff2, RSEM-EBSeq), detecting up to 50% of spiked-in transcripts.
• Performance of both analysis types improved with increased sequencing read counts.

Conclusions:

• Low-input RNA data (NuGEN v2) is suboptimal for alternative splicing detection, particularly with exon-level analysis.
• The number of input reads critically influences the performance of both splice variant quantification and exon-level analyses.
• Ribosomal RNA depletion protocols exhibited lower sensitivity for detecting splice variants due to reads mapping to non-coding transcripts.