Pros and cons of Illumina HiSeq and Next Seq

Question

[NB]

Hello,

In choosing a Next-gen seq platform, given the choices are HiSeq and NextSeq, which one would be more benefical? I know that HiSeq is very high throughput, compared to Nextseq but Nextseq produces sequences faster. Considering our sample size would be anywhere between 1500-3000 and mostly our experiments would be WGS, WES and a bit of functional work as well (methylation, RNA-seq), does anyone have any views from a data-analysis perspective as to pros and cons of the 2 systems ?

Many thanks

[NB]

Thank you for all your comments, but I wanted to know views more on data-analysis rather than the actual sequencing, as in once the sequences are produced, how good is the data, false positives generated, % of reads mapped, the coverage, how easy or difficult is it to handle for variant calling ?

[14134125465346445]

updated my answer

Answer 1

[14134125465346445]

There have been a couple of reports about data quality for PCR sample preps vs PCRFree preps as well as reports of the NextSeq platform on seqanswers comparing it to HiSeq V3. There is also another report (Slideshare) comparing HiSeq V4 to HiSeq V3.

Overall, the trend in those reports is that the data quality is, in order of best to worst:

Illumina PCRFree > Illumina Nano PCR >> Other PCR sample preps

Sample prep is the biggest factor for data quality at the moment, then followed by instrument differences:

HiseqV4 PCRFree > HiseqV3 PCRFree > NextSeqV1 PCRFree

There is one sweet-spot for the NextSeq which was reported by Dale Yuzuki, who works for Thermo (Ion Torrent), which is the low price per 100M reads for 1x75bp runs in the NextSeq, which are commonly used in RNA tag counting experiments:

http://www.yuzuki.org/clarifications-ion-torrent-pii-nextseq-500/

In terms of data management and analysis, the NextSeq is highly integrated with Illumina BaseSpace. From the design of the experiment, barcoding of the samples, uploading of the resulting read sets up to the cloud, and then running of analysis in BaseSpace apps and sharing the results with collaborators.

The HiSeq machines are gaining data integration, but it is not as seamless as the NextSeq at the moment (2014-11-28).

If you don't want to upload your datasets on the BaseSpace private cloud, you can use BaseSpace OnSite, which is reportedly as good for the NextSeq as it is for the HiSeq.

In terms of throughput, find the relevant information below:

A run on the HiSeq with V4 chemistry 2x125bp takes 6 days and produces between 0.9-1Tbp of data:

http://systems.illumina.com/systems/hiseq_2500_1500/performance_specifications.html

At 15-20x per lane, and 16 lanes per run, that's 8 WGS 30-40x 2-lane genomes every 6 days, and 8*50 = 400 WGS per year not running it on Sundays and 2 weeks holiday period.

For exomes, multiply the 400 WGS by 12, and it gives you 4800 WES per year.

If you do half and half WGS and WES, you can do 200 WGS and 2400 WES per year with a HiSeq using V4 chemistry.

For RNA profiling, multiply 400 by 24, and it gives you 9600 RNAseq per year.

If you can do 200 WGS, 1200 WES and 2400 RNAseq per year with a HiSeq using V4 chemistry.

http://systems.illumina.com/systems/sequencing-platform-comparison.html

[NB]

Thank you..

Answer 2

[Brian Bushnell]

If you want to sequence 3000 whole human-size genomes to 30x coverage, that's ~270 terabases. The configurations and loading of HiSeqs are highly variable but if you could produce 1 Tbp per run, and a run took 2 weeks, that would take 10 years on 1 HiSeq if it ran at 100% capacity, which they don't.

So, I don't see how you could possibly consider MiSeq for something of that scale. But it may depend on your specific mix of experiments. If you actually are going to do more like 1500 libraries to study the methylation of a single gene, the amount of output would be drastically less. You should try to project how many to bp of output you plan to generate; without a rough estimate, you may as well flip a coin.

Oh... and another advantage of MiSeq is that it currently generates longer, higher-quality reads.

Answer 3

[mikhail.shugay]

I believe on should approach the problem from another angle. The typical yield of HiSeq is 200-300M reads per lane, while MiSeq yields 20-30M reads. So basically there is a huge lane between a run yield of those two instruments. Due to the ability to multiplex samples, some experimental designs will surely fit in this gap. Therefore, in my humble opinion, one of the pros of NextSeq is its ability to fill this yield range.