upon genome-wide search for pvt1-like transcripts (expression levels, intron retention average across hundreds of samples, intronic miR-200 seed sites), and assessing splicing-based models for more candidates, PVT1 emerges as a top candidate with chromatin-associated, intron-retained ceRNA potential

Mechanistically, PVT1 contains several miR-200 seed sites, perfect 7-mers, located within its intronic regions; artificial splicing enhancement of PVT1 with dCasRx-RBM25 (doi.org/10.1016/j.molcel.2024.05.028) in MCF-7 reduced PVT1 intron retention and altered expression of miR-200 target genes (6/n)

Causal inference analysis using tumor-specific somatic mutations (WGS/WXS) clustered near PVT1 splice sites and associated with perturbed PVT1 splicing uncovers specific terms and miR-200 target genes enriched in datasets best explained by the causal model (among eight alternative models tested) 5/n

PVT1 has several alternative transcript isoforms detected in breast cancer samples (quantified with Salmon doi.org/10.1038/nmeth.4197), yet splicing-based predictive models outperform transcript-expression-based models in predicting distal gene expression (4/n)

We extracted splicing efficiency at PVT1 3' splice sites across hundreds of breast cancer data (TCGA) and built machine learning models to predict gene expression in genome-wide in silico screens; we find specific terms and miR-200/205 target genes enriched among high-confidence predicted genes (3/n

In previous work we measured chromatin dissociation dynamics of nascent RNAs, and profiled co- and post-transcriptional processing doi.org/10.1016/j.cels.2023.09.005; Here, we frame PVT1 as a chromatin-associated lncRNA with inefficient processing;