Benchmarking RNA-seq with the Quartet and MAQC reference materials to establish best practices for accurate alternative splicing analysis

Abstract

Previous limited characterization of short-read RNA-seq (srRNA-seq) accuracy in alternative splicing (AS) analysis due to methodological diversity and lack of reference standards, has left unclear how to achieve optimal performance—an issue increasingly critical with the rise of long-read sequencing. To address this, we conduct a large-scale reference-based benchmarking study across 42 laboratories and 207 analysis pipelines leveraging the Quartet and MAQC reference materials. Here, we show that high data quality and depth improved the accuracy of splice junction detection, as well as isoform- and event-level quantification and differential analysis. Best practices for experimental and bioinformatic design are identified, with optimal pipelines achieving Pearson and Matthews correlation coefficients of 0.79 and 0.68 for isoform-level quantification and differential analysis, and 0.41 and 0.41 for event-level analyses, respectively. This corresponds to improvements of 0.21–0.45 and 0.51–0.67 at the isoform level, and 0.09–0.27 and 0.16–0.34 at the event level relative to the poorest-performing pipelines across laboratories. Beyond technical workflows, low expression or coverage and high compositional complexity represent general constraints on accuracy. Collectively, this study provides practical guidance for maximizing AS profiling accuracy with existing methodologies, contributing to effective srRNA-seq application in RNA splicing research. Alternative splicing detection performance by srRNA-seq lacks systematic benchmarking. Here, the authors assessed isoform- and event-level performance across 42 laboratories and 207 pipelines using large-scale reference datasets, identified key factors, and provided best practices.