Paean: A Unified and Efficient Transcriptome Quantification System Using Heterogeneous Computing.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA sequencing (RNA-seq) data analysis demands high computational performance.
• Existing CPU-based pipelines are resource-intensive and slow for large datasets, limiting gene expression and alternative splicing analysis.
• There is a need for efficient computational methods to handle the growing volume of RNA-seq data.

Purpose of the Study:

• To develop a high-performance parallel computing system for rapid RNA-seq data analysis.
• To integrate efficient algorithms for calculating gene expression levels and alternative splicing ratios.
• To validate the accuracy, reliability, and applicability of the developed system on massive datasets.

Main Methods:

• Developed Paean, a parallel computing system utilizing GPU-CPU platforms.
• Integrated two efficient, highly parallel algorithms into a unified pipeline.
• Tested Paean on various computational platforms and large-scale datasets, including the TCGA dataset.

Main Results:

• Paean achieves comparable quantification precision to existing methods.
• Paean is over 100-fold faster than existing methods, requiring significantly less computation resource.
• Successful analysis of the complete TCGA dataset (10,866 samples) in 3,400 minutes demonstrates excellent performance on massive datasets.

Conclusions:

• Paean provides a rapid and efficient solution for analyzing large RNA-seq datasets.
• The system facilitates the identification of tumor neoantigen candidates from RNA splicing events, with applications in cancer immunotherapy and precision medicine.
• Paean is open-source, promoting accessibility and further development in the field.