1956P - Integrative analysis of 26848 human transcriptomes reveals cancer-defining transcriptional architecture at the isoform resolution

Background

Alternative promoters and splice sites enable most human genes to generate multiple transcript isoforms, significantly increasing phenotypic complexity. Large-scale tumor RNA-seq analysis has demonstrated widespread splicing aberrations and alternative promoter activation in cancer, implicating isoform dysregulation as a crucial factor in carcinogenesis.

Methods

By integrating RNA-seq data from TCGA, PCAWG, and GTEx projects, we reconstructed individual transcript sequences expressed in more than 26000 cancer and normal samples, elucidating a comprehensive panorama of alternative isoform utilization across 39 tissues and 38 cancer types.

Results

Our isoform-centric analysis reveals that (1) global transcriptome architecture is distinct across tissues, between cancer and normal samples of the same tissue origin, and between cancer molecular subtypes—even at sub-gene resolution; (2) alternative promoter usage and splicing together contribute to ∼40% of transcriptome heterogeneity across tissues and 50-60% within tissues and cancer types; (3) cancer frequently increases the usage of aberrant tumor suppressor gene isoforms lacking protein-coding potential; (4) differentially used isoforms hold significant prognostic value in various cancers, with aberrant transcripts from NRG1, CAMTA1, and LRP1B demonstrating the strongest association with survival; and (5) co-occurring differential splicing events within single RNA molecules may offset each other's frame-shifting effects.

Conclusions

Our findings indicate cancer frequently dysregulates alternative isoform usage, perturbing key oncogenic pathways. This highlights the value of isoform-resolution transcriptome profiling for improved individualized cancer diagnostics and treatment strategies.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Korea Advanced Institute of Science and Technology.

Disclosure

All authors have declared no conflicts of interest.