Thread by @tuuliel

I'm excited to share our new preprint, led by my postdoc @g_dafni and @beryl_bbc in collaboration with Garrett Garborcauskas, @dgmacarthur, @nanopore and others: Transcriptome variation in human tissues revealed by long-read sequencing.

Thread 👇
biorxiv.org/content/10.110…

We introduce the largest long-read transcriptome sequencing data from human tissues, from GTEx samples. The ONT technology provides an exciting opportunity for improved analysis of the real biological unit of the transcriptome: transcripts - not genes or splice junctions. 2/8

We identified about 100K novel transcripts that validate well in proteome data and are enriched for tissue-specific expression and capture biological signals. This is an important extension of human transcript annotations and shows promise for the future. 3/8

To analyze genetic effects by allele-specific expression and transcript structure analysis, we first developed a new tool, LORALS, to extract these data from long reads in a scalable manner, accounting for specific properties of these data. LORALS is available in github. 4/8

Allele-specific transcript structure captures multiple types of transcript events, and we demonstrate how the additional resolution into the exact molecular events especially in the UTRs captures novel features of genetic regulatory effects, including eQTLs and sQTLs. 5/8

We see that eQTLs actually affect not just expression levels but also 5' structure. By knocking down an RNA-binding protein PTBP1, we show how change in cellular splicing machinery manifests in changes splicing regulation and the molecular function of genetic variants. 6/8

Finally, we demonstrate potential in better interpretation of rare variants, both by improved gene annotation and by describing how rare variants disrupting transcript structure manifest in long-read transcriptome data. 7/8

The data will hopefully be an important resource for the genomics community and complements the short-read GTEx data set. In addition to the biological discoveries we make, the analyses demonstrate a lot of potential for these data types. Looking forward to your feedback! /fin