A ratio-based framework using Quartet reference materials for integrating long- and short-read RNA-seq

Abstract

Long-read RNA sequencing (lrRNA-seq) enables full-length transcript profiling but is confounded by technical batch effects that compromise quantification and prevent data integration across platforms, protocols, and laboratories. The lack of a transcriptome-wide biological ground truth has hindered objective benchmarking. To address these dual challenges, we leveraged certified Quartet reference materials to generate one of the largest multi-center lrRNA-seq resources to date: over one billion long reads from 144 libraries across four PacBio and Nanopore protocols in four independent laboratories. We first establish that ratio-based quantification against built-in reference samples effectively removes technical noise, revealing underlying biological signals. We then constructed the first ratio-based reference datasets for full-length transcripts-- comprising 10,218 isoforms and 6,032 alternative splicing (AS) events--and orthogonally validated them with RT-qPCR. Finally, a comprehensive benchmark using these ground truths reveals that a hybrid strategy integrating long- and short-read data (hybrid-seq) achieves the highest quantification accuracy for both isoforms and AS events. Our work provides a foundational framework and resource for evaluating lrRNA-seq technologies and accelerating the standardization of full-length transcriptomics for research and clinical applications.