There are two major NSD3 isoforms expressed, long (aa 1-1437) and short (aa 1-645, differing in sequence from 620-645). The short isoform lacks a methyltransferase domain and, perhaps surprisingly, is the form attributed to NSD3’s role in acute myeloid leukemia (AML) . I’ve previously used the TCGA-LUSC data-set  to test for differential expression based on the abundance of NSD3 transcripts at the gene-level. Next, I was interested in using the isoform-level expression data from Firebrowse to look at how different isoforms may be regulated. Is expression correlated or are there more complex regulatory mechanisms that control splicing decisions?
Firstly, I found that the short isoform is expressed roughly 10-fold over long isoforms, making it the dominant isoform in the LUSC data (Figure 1 – B). Next, when comparing transcript levels of long vs short, I observed a significant moderately positive correlation, suggesting that the isoforms are likely co-regulated. I hope the analysis will be helpful in framing the shared and separate functions of the two isoforms uncovered in future experiments.
As always, the complete analysis & code can be found on zenodo.
Acknowledgement: The results shown here are in whole or part based upon data generated by the TCGA Research Network: http://cancergenome.nih.gov/.
1. Shen C, Ipsaro JJ, Shi J, et al. NSD3-short is an adaptor protein that couples BRD4 to the CHD8 chromatin remodeler. Molecular cell. 2015;60(6):847-859. doi:10.1016/j.molcel.2015.10.033.
2. Weinstein JN, Collisson EA, Mills GB, et al. The Cancer Genome Atlas Pan-Cancer Analysis Project. Nature genetics. 2013;45(10):1113-1120. doi:10.1038/ng.2764.