TUSCO: benchmarking transcriptome reconstruction with endogenous single-isoform controls

Area of Science:

• Genomics
• Transcriptomics
• Bioinformatics

Background:

• Long-read sequencing (LRS) offers comprehensive transcriptome analysis but faces challenges like sequencing errors and biases.
• Existing benchmarking tools (BUSCO, SIRVs, ERCCs) have limitations in accuracy and real-world sample complexity representation.
• Simulation algorithms struggle to fully capture the complexity of actual biological samples.

Purpose of the Study:

• To introduce the Transcriptome Universal Single-isoform COntrol (TUSCO) benchmarking framework for LRS transcriptome analysis.
• To establish a reliable internal ground truth using a curated gene set lacking alternative isoforms.
• To enhance the accuracy and reliability of transcriptome reconstruction quality control.

Main Methods:

• Developed the TUSCO framework utilizing a curated gene set of single-isoform genes as an internal ground truth.
• Quantified precision by detecting reconstructed transcripts deviating from reference annotations.
• Assessed sensitivity by verifying detection completeness in human and mouse samples.
• Created a TUSCO-novel challenge to evaluate the reconstruction of unannotated isoforms.

Main Results:

• TUSCO metrics provide accurate and reliable transcriptome benchmarking without external controls.
• The framework effectively addresses limitations of existing methods like BUSCO and spike-ins.
• Demonstrated improved quality control standards for LRS transcriptome reconstruction.

Conclusions:

• TUSCO offers a robust solution for benchmarking LRS transcriptome data.
• The framework enhances the reliability of transcriptome reconstruction and quality control.
• TUSCO significantly advances the field of transcriptome analysis using long-read sequencing.