Introduction: A New Lens on Gene Expression at ASHG 2025
The ASHG 2025 meeting in Turin spotlighted a transformative approach to understanding how genes express themselves in health and disease. Researchers from the University of Chicago and Columbia University presented Torino, a novel computational workflow that decodes transcript structures and expression levels directly from RNA sequencing read coverage, bypassing the limitations of traditional annotation-dependent methods. This breakthrough promises to illuminate the hidden diversity of RNA processing and its links to disease phenotypes.
What Torino Does
Torino tackles a central challenge in transcriptomics: many disease-relevant isoforms remain unannotated yet actively expressed. Conventional RNA-seq analyses rely on existing transcript catalogs, which can overlook novel isoforms generated by alternative splicing, intron retention, and alternative polyadenylation. Torino uses Poisson non-negative matrix factorization with spatial smoothness priors to infer latent transcript structures from read coverage data alone. In doing so, it enables transcriptome-wide discovery of previously unknown, functionally important isoforms and their regulatory patterns.
Key Findings Across the GTEx Resource
Applying Torino to the Genotype-Tissue Expression (GTEx) project, which encompasses 2,128 samples across 19 tissues, yielded striking results. The workflow successfully recovered 18,813 GENCODE isoforms spanning 15,232 protein-coding genes. More importantly, it unveiled extensive unannotated diversity, including:
– Over 10,000 novel cassette exon events
– More than 53,000 novel intron retention events
– About 8,013 alternative polyadenylation (APA) events
Many of these events display tissue specificity, suggesting they play meaningful roles in defining cell identity and function. The discovery of such widespread unannotated isoforms reframes how we understand gene regulation in different biological contexts.
Genetic Control of RNA Processing (isoQTLs)
By leveraging Torino-inferred isoform abundances, the study identified a median of 2,829 isoform QTLs (isoQTLs) per tissue, underscoring pervasive genetic influence over RNA processing. Colocalization analyses linked 815 disease-associated variants with isoQTLs across multiple GWAS traits, pointing to unannotated splicing events as potential hidden drivers of disease risk. This connection between genetic variation, RNA processing, and disease opens new avenues for understanding complex genetic architectures beyond traditional gene-level analyses.
Implications for Disease Research
Incorporating unannotated isoforms into genetic and functional analyses has multiple implications:
– It reframes disease models by highlighting RNA processing variants as potential causal factors.
– It improves the interpretation of GWAS signals by mapping them to specific, alternatively processed transcript isoforms.
– It provides a richer landscape for exploring tissue-specific pathophysiology, which is crucial for diseases with strong neurobiological or systemic components.
Insights from Brain Data and Alzheimer’s Disease
When Torino was applied to 1,193 brain samples from the Alzheimer’s Disease Functional Genomics Consortium, a global increase in intron retention emerged as associated with Alzheimer’s disease (AD) diagnosis and Braak stage, a measure of neurofibrillary tangle pathology. Notably, splicing abnormalities were observed in AD risk genes PTK2B and APBB3, suggesting that RNA processing defects may contribute to disease progression and could serve as potential biomarkers or therapeutic targets.
Future Directions and Impact
Torino demonstrates the feasibility of decoding the regulatory genome across additional tissues, species, and complex diseases. As researchers integrate Torino with broader datasets, we can expect a deeper understanding of how transcript isoform diversity shapes cellular identity and disease mechanisms. The approach also holds promise for refining precision medicine, by identifying disease-relevant isoforms that could serve as diagnostic indicators or treatment targets.
Conclusion: A New Era in Transcriptomics
By directly reading transcript structures from read coverage, Torino bridges RNA processing complexity with genetic architecture and disease. The ASHG 2025 presentation marks a pivotal step toward a more complete map of the human transcriptome, revealing that the disease-relevant RNA landscape includes many unannotated isoforms with tissue-specific roles. As this field advances, Torino’s framework could redefine how we study gene expression, RNA processing, and their collective impact on human health.
