a bit off-topic, but is anyone aware of thorough benchmarks between 16s Illumina and PacBio profiling? What are the gains and is it worth the higher price or should I directly go for metagenomics?
Great question! I’ve never used PacBio personally but a colleague of mine recently cited lack of user support for analyzing PacBio type data as the most challenging part so far. and suggested that with proper planning the cost itself was not that different than short-read sequencing. Historically the big disadvantage for long reads (aside from the cost) was the higher error rates but these are being resolved very quickly, for example @benjjneb’s awesome new pre-print showing PacBio’s CSS method paired with DADA2 inference giving near-zero error rates mock communities! He may have much better insight on the matter as well. Another pre-print with similar results with Nanopore also came out recently that might be of relevance.
Another paper worth reading (though I actually haven’t gotten around to reading it thoroughly myself) is this one by Whon et al. 2018 which does some technology comparisons and seems to highlight the need for better reference databases even with long-reads.
As for going for going directly for metagenomics, I think that bit is really more dependent on the question. If only bacterial community composition is the end goal then I think 16S (short or long reads) will more than suffice, but if you need more in depth profiling i.e. inter-kingdom relations, functional capacity etc. then I think shot-gun sequencing is still your best bet. And depending on whether or not the target community is already well defined, you may utilize shallow shotgun sequencing which significantly reduces the price of it to short-read sequencing levels.
Anyways, I know this isn’t exactly what you were looking for but maybe this gets the conversation going! Looking forward to hearing the experts’ view on the matter.