Thread by @AlbertVilella

Reviewing the Oxford @nanopore presentation now available to replay. Two highlights/thoughts that I didn't tweet about before:
1) Spike-in for re-training
2) Strand-specific DNA modifications

1) Looking at the kind of stuff presented, where temperature settings, salt, etc. are being more available to end-users, there is an opportunity here to retrain models and improve basecalling for specific datasets.

On a much simpler level, Illumina sequencer did and still do some of that: one would need to spike-in a certain amount of PhiX for the machine to calibrate itself. For low-complexity applications, like 16S sequencing, this could be as high as 25-50% of the reads/throughput.

Even though not explicitly mentioned by @The__Taybor in the latest presentation (as of 2022-04-08), this maybe something to think about: spike-in certain known sequences, and together with the measured temperature/salts/etc of the run, then use that to re-train the models and ...

... then basecall at higher accuracy. One of my obsessions with #NGS is #epigenomics, and thus improvements on accuracy of 5mC/5hmC/5fC/etc are in my opinion move the needle forward in #LiquidBiopsy and other important applications.

2) Strand-specific base modifications:
Even though Oxford @nanopore has already surpassed bisulfite sequencing with their 5mC base calling, there is another aspect of this that I want to highlight.

The fact that Duplex calling of base modifications works on the basis of a non-PCRed molecule (double-stranded DNA), one guarantees that both strands came directly from the sample, thus when sequencing both strands and calling base modifications...

... the modified bases are naturally strand-specific. This can have important applications, as the order and tempo with which the base modifier enzymes perform inside the cells matters in understanding cell states.

Now, even though there are lots of scientific publications on strand-specific base modifications and their link to cell biology and cell development, I think the field is not as mature in diagnostics as it is for the non-strand specific use of #epigenomics for diagnostics.

Unless someone with more knowledge can correct me, applications such as the ones GrailBio and Guardant Health are using are non-strand specific, either by design or by technology limitation. And none of them uses yet a combination of 5mC / 5hmC, which relate to each other.

And here is where I think there is an opportunity for either Oxford @nanopore or adaptations of current methods using short-read 4-base sequencing to do both profiling of 5mC + 5hmC but adding the strand-specific readout that comes naturally from nanopore Duplex reading.

Still something to mull on further, but I believe that the signal is there to be mined.