Blessy: enabling differential analysis of phased protein domains | LC 25

Biography

Dr Nadia Davidson leads a bioinformatics research group at the Walter and Eliza Hall Institute, Australia. She received her PhD in Particle Physics from the University of Melbourne in 2011, before transitioning into bioinformatics. Her research focuses on developing computational methods to analyse short and long RNA sequencing data. She has authored more than 12 highly utilised open-source software packages across research areas, including fusion gene detection in cancer (JAFFAL), transcriptome analysis in non-model organisms, and long-read single-cell analyses (Flexiplex).

Abstract

Differential analysis through RNA sequencing has revolutionized biology, unveiling the genetic processes that govern development and disease, and revealing novel treatment targets. While gene expression has traditionally been examined, transcript-level analysis allows isoforms with differing — or even opposing — functions to be assessed independently.

Oxford Nanopore Technologies sequencing is particularly well-suited for this purpose, as long reads can be unambiguously assigned to transcripts. However, transcript-level analyses suffer from reduced statistical power compared to gene-level, a limitation compounded in sparse single-cell data. In addition, interpreting the impact at the protein level remains challenging and generally requires additional analyses.

To address these limitations, we have developed a new approach that groups transcripts with similar biological function, specifically by phasing protein domains. Protein domains are highly conserved coding regions within genes that are predicted to have distinct roles, such as binding to DNA or localising to the transmembrane region.

Blessy is a new open-source R software that we have developed, which identifies the protein domain combinations (DoCos) within each transcript, and aggregates transcript counts based on these combinations. We validated Blessy using Oxford Nanopore Technologies bulk and single-cell sequencing data, demonstrating its ability to identify isoform-switching events that result in changes in protein domain usage.

Notably, DoCos captured known changes in cancer with higher statistical power than traditional transcript-level analyses and with simplified interpretation. Blessy is versatile, working with custom domain annotations and even other genomic features, making it a powerful and practical tool for long-read transcriptome analysis.

Blessy is available at https://github.com/DavidsonGroup/longhaul/.