1
$\begingroup$

Any comparable alternatives to linked read sequencing?

I read that 10X Genomics discontinued its linked-reads technologies:

Discontinuation of Linked-Reads At 10x Genomics, we are committed to enabling your research and providing you with >the best experience to make scientific discoveries. Earlier this year, as a part of >our product life cycle management, we notified you about refinements to some of our >on-market product offerings.

As of June 30, 2020, we discontinued the sale of our Chromium Genome and Exome >product lines. We will not be releasing new versions of these products.

We thank you for your partnership and are committed to providing the highest-quality >technical and software support.

Please contact your Sales Executive, Field Application Scientist, or email our >Support Team at support@10xgenomics.com with any questions.

See here on the 10X Genomics homepage.

This source assumes the product was discontinued due to...

infringed patents held by Bio-Rad Laboratories...

But when searching for "linked-reads" and "Bio-Rad" I cannot find anything. Having worked with ddPCR I assume 10X Genomics problem was the oil emulsion but not the linked-reads approach?

Is there an alternative platform doing what 10X Genomics linked-reads was able to accomplish? I am specifically interested in its role in genome assembly/scaffolding as in the Vertebrate Genome Project (in bold):

The current pipeline (Figure 1) to meet the 3.4.2.QV40 phased metric with the fewest errors currently achievable consists of a combination of the following approaches:

  1. 60X PacBio long-reads for an initial phased contig assembly (30X/haplotype);

2. 68X coverage of 10X Genomics-linked reads for intermediate-range scaffolding and further phasing (34X/haplotype);

  1. 80X Bionano optical maps to correct scaffolding errors and for further scaffolding;
  1. 68X Hi-C linked reads for long-range scaffolding; PacBio Jelly algorithm to fill gaps using long-reads;
  1. 10X Genomics Illumina short-reads for base-call accuracy polishing; and

Thanks for your insights!

$\endgroup$
4
  • 1
    $\begingroup$ What is the specific goal/need that you have that linked-reads satisfied? Is it related to your previous question here? biology.stackexchange.com/questions/101425/… $\endgroup$ Sep 8 at 19:11
  • 1
    $\begingroup$ This is a question about bioinformatics software, so you will have a better chance of getting an answer on SE Bioinformatics. $\endgroup$
    – David
    Sep 8 at 19:11
  • $\begingroup$ @MaximilianPress no I have been reading about the pipeline of the Vertebrate Genome Project. They have been using 10X genomics' linked-reads. I wonder how the pipeline might change in the future. rockefeller.edu/research/vertebrate-genomes-project/… I thought someone might know. Or immediately point out a reason why it might have become obsolete given other technologies on the market. $\endgroup$
    – ilam engl
    Sep 9 at 10:40
  • $\begingroup$ First, I was just expressing an opinion on the SE group most likely to provide an answer to a valid question. I may or may not be right. As far as ddPCR my guess is that this list might be better as it’s not particularly Bioinformatics, but you’d be expected to research the topic first. Don’t worry too much, if people think you have posted to the wrong list the Moderators can move or help you move your question. $\endgroup$
    – David
    Sep 9 at 12:42
4
$\begingroup$

Maximillian provided some good information, but I thought I'd add some practical considerations:

  • All of the long-read sequencing platforms provide quite contiguous assemblies (scaffold N50s up to 40Mb or so), but those are definitely not chromosome-scale. No matter which platform you use, you will still need a "bigger-scale" method to get to the end. This can be Hi-C, strand-seq or optical mapping.

  • PacBio Hifi reads are great in that they are high-accuracy and allow high contiguity, but, at least in my experience, you will pay 2-4 times as much to get a comparable result to what you could get with 10X linked reads.

  • Nanopore can give you a very contiguous assembly with 20-30X coverage (ie about a PromethION flowcell's worth, roughly equivalent in cost to a 10X run on Illumina), but as Maximillian noted, requires polishing with short reads to get the base error corrected. (Note: Nanopore's new Q20+ sequencing is looking promising for this, but costs about twice as much per Gb. It's still cheaper than PacBio though.)

  • So with that said, it's extremely tragic that the cheapest method for making a high-accuracy, one-step-from-chromosome-scale assembly was killed as collateral damage in a patent war.

And then to answer your question:

  • MGI have been prototyping a method called stLFR that was supposed to be a replacement, but that seems to still be a work in progress, and also may have been shelved during the pandemic as they shifted towards supporting COVID sequencing.
  • TellSeq is another possible option, although I can't comment yet on how well it works in practice.
$\endgroup$
3
$\begingroup$

Update:

Answer was transferred to Kieran O'Neill's response, which I think is appropriate. I am not completely convinced that these technologies are necessary, but I am also not as familiar with them.

Original answer

The VGP used 10X linked-reads for 2 purposes, as I have updated your question to reflect:

  1. intermediate-range scaffolding
  2. further phasing

I would argue that both of these purposes have been to some extent rendered unnecessary by advances in technology. VGP protocols seem to be frozen at this point to ensure comparability I guess, but they are now somewhat outdated. For example, the VGP also calls for PacBio CLR technology, which is at this point not preferred for genome assembly. Instead, most would use PacBio HiFi reads, for which an extremely strong suite of tools such as HiFiAsm already exists. HiFiAsm and HiFi tools in general do an excellent job of phasing.

The other component is Hi-C-seq, which provides both intermediate-range and long-range scaffolding information to generate chromosome-scale scaffolds from contigs. The only reason to use intermediate range scaffolding is to simplify the Hi-C scaffolding problem, which is unnecessary with a higher quality of assembly such as that provided by HiFi. Hi-C can also be used for long-range phasing, and is at this point built into HiFiAsm, uniting the two data types.

A similar argument can be made that BioNano is no longer necessary.

In fact, HiFiAsm can use only these two data types (HiFi and Hi-C) in concert to create effectively chromosome-scale diploid human genome assemblies. In my experience it still requires some manual curation to get to full chromosome scale, but the results are still very impressive.

I would say that, at this point, steps 2, 3, and possibly 5 of the VGP are no longer necessary, and step 1 should be replaced by HiFi.

A somewhat lower-cost workflow might be as follows:

  1. 60X Nanopore long read sequencing and assembly. For a recent tool see here.
  2. polishing nanopore assembly with nanopore reads using e.g. medaka, possibly diploid phasing.
  3. 100X Hi-C sequencing for scaffolding assembly using Salsa2 or RagTag.
  4. Possibly polishing with high-coverage Illumina shotgun.

Additionally, there are a number of service providers such as DNANexus, Phase Genomics, DoveTail Genomics, etc. that will do some or all of these steps for a price. While I was an employee at Phase Genomics, I contributed to a guide for performing genome assemblies that is now almost as outdated as VGP but may still have useful information.

Full disclosure: former employee of Phase Genomics, hold some equity in the company still.

$\endgroup$

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.