Multi-sample, multi-platform isoform quantification using empirical Bayes

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Accurate RNA isoform quantification is vital for understanding gene regulation and disease.
• Short-read (SR) sequencing has limitations in resolving transcript ambiguity.
• Long-read (LR) sequencing reduces ambiguity but has high error rates and lower throughput.

Purpose of the Study:

• To develop a novel method for enhancing transcript quantification by integrating SR and LR sequencing data.
• To leverage multi-sample learning to improve accuracy and reproducibility in RNA isoform abundance estimation.
• To address the limitations of existing methods in quantifying low- and moderate-abundance isoforms.

Main Methods:

• Introduced JOLI, a hierarchical model utilizing multi-sample learning for joint SR and LR data integration.
• Applied an empirical Bayes framework to learn shared priors across multiple samples for consistent inference.
• Benchmarked JOLI against single-sample methods using simulated and real RNA-seq datasets.

Main Results:

• JOLI demonstrated improved ranking consistency, proportional agreement, and estimation accuracy compared to single-sample methods.
• In simulations, JOLI improved Spearman correlation by 9.8% for LR and 7.7% for SR data.
• JOLI showed significant improvements in quantifying low- to moderate-abundance isoforms and enhanced reproducibility.

Conclusions:

• JOLI effectively leverages multi-sample learning and joint SR/LR data integration for robust transcript quantification.
• The method offers superior accuracy and reproducibility, particularly for challenging low- and moderate-expression isoforms.
• JOLI performs competitively with state-of-the-art approaches, highlighting its utility in transcriptomic studies.