We identify novel interferon stimulated transcripts and genes. We also found elongation of poly(A) tails of genes encoding ribosomal proteins in response to interferon stimulation (but not influenza infection), indicating this may be an important part of the innate immune response.

The Ferret is widely considered as the gold-standard animal model for influenza infection. Ferrets are infected with the same strains of Influenza that infect humans, and show similar symptoms (sneezing, fever)

As well as enriching for target host transcripts, we also show that it can be used to enrich for viral transcripts in infected cells, in this case SARS-Cov2

And can be used investigate differential isoform usage

The method provides ~10 fold enrichment across a range of different expression levels

I should add that we decided to not report RNA modifications in this preprint as we use outdated R9.4 chemistry, and there are many modifications and many tools, which often given conflicting results. But this is also an exciting area for biomarker discovery.

Overall we think there is a lot of potential for finding new biomarkers of disease using native RNA sequencing. One thing to note is that nanopore direct RNA sequencing is quite a bit lower throughput than cDNA sequencing, so its still more expensive.

Even though we had a relatively small sample size (12), we also looked to find potential polyadenylation biomarkers of bacterial vs viral infection. Left panel shows normal gene expression count based biomarkers (reflecting what we saw with larger cDNA-seq study), right panel shows polyA biomarkers

After filtering likely artefacts, we found a modest number (594) of new transcript isoforms, including 9 new isoforms for IGLL5. We also looked for differential transcript usage, and found 4 genes with statistically significant differential transcript usage between bacterial and viral infection

We asked whether polyadenylation varied by pathway/function. This plot shows polyA tails of transcripts in significant GO molecular functions in a gene set enrichment analysis for median tail length (excluding mitochondrial genes)

Firstly we wanted to see if dRNA sequencing based gene counts were similar to cDNA based gene counts. The answer depends a lot on which analysis pipeline you use. Algorithms which use expectation maximisation seem to be better correlated (e.g. nanocount with kallisto)

To do this we re-sequenced left-over RNA from the same extraction for 12 samples which had been previously sequenced as part of a larger sepsis biomarker study using Illumina cDNA sequencing (www.thelancet.com/journals/lan...