7
$\begingroup$

PacBio provides longer read length than Illumina's short-length reads. Longer reads offer better opportunity for genome assembly, structural variant calling. It is not worse than short reads for calling SNP/indels, quantifying transcripts. Sounds like PacBio can do whatever Illumina platform can offer.

Would that be any reason for going with Illumina if PacBio's long reads can do everything and more?

$\endgroup$
  • 2
    $\begingroup$ PacBio used to have high error rates and the expensive libraries with low throughput. In terms of transcriptome, its not quantitative, needs to generate several libraries with different fragment lengths, again not quantitative, captures only abundant transcripts, misses LncRNAs. $\endgroup$ – geek_y Jan 31 '19 at 6:29
  • 3
    $\begingroup$ IMHO not all NGS questions are about bioinformatics, @gringer, and not all bioinformatics has to do with NGS. So I kindly disagree with your statement about "Bioinformatics includes NGS". This question is about the specs of different platforms, and could be asked on a forum such as SEQanswers. $\endgroup$ – benn Jan 31 '19 at 9:53
  • 2
    $\begingroup$ This is a relevant recent preprint: biorxiv.org/content/10.1101/519025v2 TLDR: CCS improves quality, until recently at a cost of reduced length. But still a quite expensive approach. $\endgroup$ – Wouter De Coster Jan 31 '19 at 10:00
  • 3
    $\begingroup$ @benn Analyzing NGS data is very much a part of bioinformatics. By extension, the generation of NGS data is also on topic here since it is very relevant to the job of the bioinformatician who will analyze them. I don't understand why you would consider this off topic. We've even had a meta discussion about this sort of thing and the consensus was clear: biological questions are on topic when they are relevant to bioinformatics. $\endgroup$ – terdon Jan 31 '19 at 10:08
  • 2
    $\begingroup$ The close votes claim that this question is opinion-based, but there are some very objective answers to the question of which sequencing platform to choose. I think the question itself could be re-worded a bit, but it's definitely an on-topic discussion. $\endgroup$ – Daniel Standage Jan 31 '19 at 14:09
14
$\begingroup$

There are so many reasons why one might want to prefer Illumina over PacBio (also note that it's a false dichotomy, at least Oxford Nanopore is a competitive sequencing platform):

  • The first (IMHO and the most common reason) is still the cost of both sequencing and the instruments. Illumina can sequence a Gbp of data for \$7 - \$93. PacBio sequencing is according the same webpage \$115 per Gbp, however at our sequencing center it's ~$200. Though ONT might have already a cheaper solution. Edit, I just found a google sheat with prices that seems to be frequently updated, seems that the ratio still holds Illumina short reads ~10x cheaper than PacBio.
  • RNA-seq (i.e. analysis of a gene expression) is not possible with PacBio due to preferential sequencing of smaller fragments; shorter genes would always be shown to be more expressed. To be clear, it's possible to sequence RNA with PacBio (the keyword is iso-seq), but the analysis of gene expression is problematic.
  • It's way easier to extract fragmented DNA (concerns small non-model organisms; although recently a single mosquito was sequenced, so we can expect a further improvement)
  • other sequencing techniques as RAD-seq that allow genotyping with very little effort and cost, I have never seen anybody even considering using long reads for such genotyping
  • Genome profiling (assembly-less genome evaluation) based on kmer spectra analysis is not possible with PacBio data due to higher error rates. Conflict of interest: I am a developer of one of the tools for genome profiling (smudgeplot)
  • DNA in Formalin-Fixed Paraffin-Embedded samples is fragmented anyway (~300bp), therefore, there is no point in sequencing them with more expensive long-read technology (contributed by @mRoten)
  • I bet there will be a plenty other applications I am not aware of
$\endgroup$
  • $\begingroup$ Please add that FFPE sample DNA is fragmented (~300 nt), so Illumina short read sequencing is still has the lowest cost per base than PacBio. Even ConcatSeq (nature.com/articles/s41598-017-05503-w) only increase the read count by a factor of 5. Illumina is still the winner. $\endgroup$ – mRotten Mar 18 '19 at 18:06
  • $\begingroup$ I turned the answer to community wiki so it's easier for people to add/updates reasons for Illumina. As this is not complete not a stable-in-time answer. $\endgroup$ – Kamil S Jaron Mar 19 '19 at 9:11
5
$\begingroup$

Three reasons for Illumina:

* Much better for a large number of samples (easily handle 96 samples). 
* SNP calling is much better - much greater depth
* Hardware costs, an Illumina MiSeq machine is cheap 

PacBio SNP calling has been improved towards the standard of Illumina by DNA modification to the inputs to be sequenced but it depends on the application.

| improve this answer | |
$\endgroup$
5
$\begingroup$

Many analyses performed on Illumina machines these days require large numbers of reads. For example, most analyses in ChIP-seq, RNA-seq, ATAC-seq etc, need 10s or even 100s of millions of reads for the statistics to work out properly.

But this isn't limited to just sequencing as an assay experiments. High depths are important for things like somatic variant calling as well. Or simply sequencing of medical gene pannels, where you might only have 20 amplicons of a 500bp each, but need to do it on 100s of patients as cheaply as possible.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.