Accurate strand-specific long-read transcript isoform discovery and quantification at bulk, single-cell, and...

Area of Science:

• Genomics
• Transcriptomics
• Bioinformatics

Background:

• Long-read RNA sequencing advances transcript isoform profiling.
• Challenges exist in isoform identification and quantification due to partial transcripts and sparse data.
• Existing methods struggle with ambiguous read assignments and high transcriptional heterogeneity.

Purpose of the Study:

• To develop a unified computational framework, LRAA, for accurate isoform identification and quantification from long-read RNA sequencing data.
• To address challenges in read-to-isoform assignment and abundance estimation across diverse sample types.
• To provide robust analysis for bulk, single-cell, and single-nucleus transcriptomic data.

Main Methods:

• LRAA combines splice-graph based structural modeling with expectation-maximization optimization.
• The framework supports quantification-only, reference-guided, and de novo analysis modes.
• Benchmarking involved simulated and genuine datasets, including a novel MORFs strategy for ground truth.

Main Results:

• LRAA demonstrated superior performance over state-of-the-art methods in isoform identification accuracy, sensitivity, and quantification.
• The framework successfully resolved cell-type-specific isoform usage in immune cells.
• LRAA detected a pathogenic cryptic isoform of STMN2 in frontotemporal dementia (FTD) data.

Conclusions:

• LRAA is a robust and versatile solution for resolving transcript diversity in complex biological systems.
• The framework enhances the accuracy of isoform identification and quantification in transcriptomic studies.
• LRAA has significant utility in both fundamental research and disease-associated transcriptomic analyses.