Review: The 2017 NGS trends (part 1)
Did the year 2017 signal the end of exome sequencing? Or did it nod towards affordable population sequencing? Have we reached the $100 genome? And what is single cell genomics all about…?
In 2016 I wrote a reflective review on the NGS trends and what the future may bring, and it was one of the most popular articles I’ve written on social media sparking a debate.
So I’ve strived to do the same here and I’ve had to split it out into 2 parts (here’s part 1) due to the sheer amount of information and trends in this ever-growing industry. A slight disclaimer for this article that since writing the 2016 review I’ve joined illumina (the leader of NGS sequencing technology) but must stress that this is my independent opinion. Although a few of the trends focus on illumina technology and NGS applications, that’s simply purely because of the crazy amount of technology being pumped out of the Cambridge labs. Thanks colleagues.
My top trends of 2017:
- The end of exome sequencing
- Affordable population sequencing
- Rise of immuno-oncology
- Simpler library preparation
- The end of FFPE samples
- Larger and more complicated enrichment panels
- Increase in sequencing individual cells
1. The end of exome sequencing…?
The genome is the new exome
So you’re probably be asking yourself, why have I put the melodramatic title of ‘the end of exome sequencing’? Well, it might surprise you that it’s not such an exaggeration after all… This is one I wrote about in 2016 review and have had a twitter debate with multiple people about previously (Feel free to tweet me with your opinion afterwards). But let me explain before you click-away and take to social media to say how wrong this is….because you may be surprised!
— BarbaraJennings (@GeneticsMBBS) 23 January 2016
Living in the UK and with the advances from Genomics England (GeL), I can’t help but stress the rise of- and clinical usefulness of whole genome sequencing (WGS) above and beyond the exome. What we will find, and continue to find, is the continuing transition of demand from whole exome sequencing to genome sequencing. Why? Well, whole genome sequencing is quite simply more extensive than whole exome sequencing (WES). Okay, so you may agree in principle that it’s more ‘extensive’ by covering 100% of the genome rather than the 2% of coding regions that exome sequencing covers. But wait a minute! 98% of the genome (the non-protein coding regions) are just ‘junk’… right? What’s the point in sequencing those?
Two points here: a). Perhaps not! There’s lots of new research that shows the importance of regulatory elements in intronic regions and b). …Take a look at the table below. Although yes, it DOES sequencing a lot of regions that we simply cannot analyse currently, the ‘one size fits all’ approach of WGS allows for 1 instead of up to 20 different tests. This simply isn’t possible with exome sequencing:
|Type of Variant||SNPs and InDels (eg. EGFR T790M)||Repeat Expansions||Large structural variants (eg. BCR-ABL)||Copy Number Variants (eg. MET amplification)||Mito|
|Testing options||● Sanger sequencing
|● PCR||● FISH||● Chromosomal microarray
● Exome (ish)
|● Sanger sequencing
Table – WGS covers all these ‘traditional’ methodologies
The economics of WGS v WES
Well, I don’t think anyone would argue that WGS >WES in pretty much every way other than ‘cost’, and cost is an extremely important point, but do take into account my inverted commas on ‘cost’… Because, I’m not just talking about reagent costs. Think for a second beyond this:
- What does adopting a one-size fits all approach for multiple methods do for a lab in terms of costs savings?
- How much would this save a lab in unused reagents savings?
- How much does multi-method training of staff cost?
- …and what does it cost to a researcher/patient when you wait to batch samples?
- Finally, do the benefits of WGS still outweigh this additional ‘cost’?
Let’s find out and break this down and talk technical/costs:
An unknown secret of WGS compared to WES is, due to the way enrichment capture in WES is biased towards particular GC/AT rich regions, that WGS has greater coverage uniformity & sensitivity for SNP calling compared to equivalent exome sequencing coverage. The equivalent of an exome 95% detection at 40x is a WGS at 14x (source: personal knowledge) BUT with the added benefit of SV, non-coding region detection etc…. etc…
Take a look at the decreasing cost of WGS. Here’s a graph on the decreasing cost of WGS compared to WES. Furthermore, the promise of a $100 genome by Francis DeSouza (CEO of illumina) will undoubtedly spur people on to adopt WGS in preference to WES. I could also mention the newest advances in bead-based tagmentation library prep in WGS… but those would be spoilers of what I will talk about in the library prep part of the blog.
The clinical diagnostic yield is also much superior as shown in Willig, et al. 2015. Furthermore, with 35% of the pathogenic and likely pathogenic variants as shown in ClinVar that are not InDels or SNVs, there’s clearly much more we need to sequence. A good example of further diagnostic yield can be seen in the paper by Mallawarachichi et al., distinguishing between pseudogenes: Whole-genome sequencing overcomes pseudogene homology to diagnose autosomal dominant polycystic kidney disease.
Could WGS even surpass the array?
My final thoughts and I will leave you to decide, but the publication below is well worth a read comparing low pass WGS to arrays:
“Herein, a protocol providing CNV detection from low-pass, whole-genome sequencing (0.25×) in a clinical laboratory setting is described. The cost is reduced to less than $200 USD per sample and the turn-around time is within an acceptable clinically workable time-frame (7 days).” by Dong et al.
So, I feel that 2018 will be the year that I will find myself, happily, being a parrot and asking the same question:
“Why wouldn’t you do WGS?”
2. Affordable population sequencing
Big platforms are powering big genomics
Before I move onto my next top trend. Read these statistics:
- In 2007 it took 3 days to generate 1Gb of data. Today (2017) it now takes just 2.4 minutes. (FYI 120Gb = human genome at the ‘gold standard’ level (30x))
- 10 years ago, it took 1 year to sequence a human’s genome. The highest throughput NGS platform currently available on the market, the NovaSeq, can now do 48 genomes in just 48 hours.
- In 2003 a genome cost $3,000,000,000,000,000 to sequence (aka the genome project for $3bn). It now costs $1,000.
Crazy numbers right? In 2016 I predicted that we were aiming for a $50 genome courtesy of Genapsys. But I honestly haven’t heard anything from them since. Has anyone else? Regardless, a couple of months after my blog Francis (CEO of Illumina) announced the aspiration of the $100 genome driven by a new product launch earlier this year. Whether it’s $100 or $50 we’re aiming for, there’s been some great advances in this area this year….
2017 also saw the launch of some big NGS platforms to fuel population sequencing with some equally big promises, including the NovaSeq from Illumina and PromethION from ONT. Using the S4 flowcell (pictured in ICR’s tweet below), you can sequence 24 genomes in 48 hours, at a list price of ~$1,200/sample for WGS. Scale this up by 2x (2 flowcells on 1 platform) and voila, you have 48 genomes in 48 hours.
— The ICR (@ICR_London) 28 July 2017
So the economics of sequencing, driven by NGS, are important. I was particularly proud of my local National Health Service this year, and two key initiatives stood out for me in Europe related to population/clinical sequencing and mainstream adoption of NGS. These initiatives give a strong indication of where governments and companies will be heading in the future, a nod towards both ‘affordable’ and mass/population sequencing:
- My very own National Health Service in the UK announced that is preparing to commission whole genome sequencing as a routine diagnostic test for certain rare diseases and cancers.
- Finally, but by no means least, the French announced plans for what they’re calling ‘Genomic Medicine 2025‘. This plan aims to integrate genomic medicine within the French clinical care pathway, with the goal to sequence 235,000 genomes each year by 2020. Many would ask if this is this do-able? Well, looking at the email I received from GeL recently we’re currently on 41,045 genomes in the UK going into the start of 2018. So yes it is do-able and I expect/hope to see many more announcements similar to these.
3. The rise of immuno-oncology
NGS complements traditional protein-based techniques with additional actionable insight
Immuno-oncology is ‘hot’. Why? Well it fundamentally just makes sense, right? Engineering or enhancing your own immune system to fight off disease. But it’s not only because of its simple elegance, but also because of the big ‘bucks’ of over 240 immuno-oncology drugs currently in the pipeline.
Let’s take a deeper look:
Generalising, there’s 3 key components of immuno-oncology:
2). Immune modulation and;
3). Tumor micro-environment
In the interest of the blog I focus on #1 – potentially the most important as antigens are what the immune system uses to recognise foreignness, and begin the cascade of response. Without antigens, immunological responses are simply not possible.
So why the interest? Well, immune ‘checkpoints’ or controllers like CTLA4 or PDL-1 limit or downregulate immune response to prevent damage to normal cells and certain cancerous cells up-regulate these surface ligands to help the tumors to evade immune system control. It therefore makes sense that if there were, say a monoclonal antibody that inhibited these ligands, the host could mount an immune response of therapeutic benefit.
And if you were thinking that makes sense, to engineer, help or enhance your immune system to fight cancer, you’re right! (Because a system that has evolved for millions of years must be an advantage right?)
A monoclonal antibody ipilimumab (mAb) does exactly this. By targeting and blocking the CTLA4 ligand combination therapeutics should be more efficacious…. but in a study of this mAb, ipilimumab yielded a beneficial response rate of only 20%? While this number is low, it still represents a tremendous improvement from the previous standard of care.
But not all patients react the same.
Figure – Could NGS complement traditional IHC approaches?
IHC + RNA + TMB = The best approach?
Historically IHC has been the gold standard for…pretty much everything. This is very intuitive because one would think that expression of the protein that the therapy is targeting is essential in determining response. And yes, in this trial IHC independently predicted response.
However, from this study and others that some IHC positive patients don’t respond, and some IHC negative patients do. It is for this reason that the authors took a more comprehensive approach to biomarkers. Both RNA-Seq and mutational load, also served as independent predictors of response.
And as no surprise when the these biomarkers were combined with IHC, the best prediction of response was observed.
Is a multi-omic future closer than we think?
4. Simpler library preparation
One day, even ‘Joe Bloggs’ will be able to do NGS
Prior to 2017, preparing your samples for NGS (library preparation) was either performed one of two main ways: ‘mechanically’ via using a Covaris or Biorupter (epigenetics) that, although slow and time consuming, gave a consistent quality or through the faster ‘enzymatic’ approach of tagmentation at the trade-off of quality. However once of the biggest pain points of Nextera (enzymatic method) prep’s has been controlling insert size, where over-tagmentation can lead to short inserts and thus scientists had to train to become experts in this methodology. Examples of this include NEB Next Ultra DNA, KAPA HyperPlus and Agilent SureSelectXT HS (the latter also including UMIs). This is one of the biggest barriers for small inexperienced labs in adopting this technology without a doubt. A barrier that has now. Firmly. Been. Removed.
So what’s happened this in 2017 to address this?
Well, some scientists have been busy beavering away to produce what I’m calling the ‘Joe Bloggs’ library preparation method. Quite simply, if you have over 100ng of input DNA the protocol is so straight forward that you can have a NGS-ready library in <4 hours. Take into account previously methods required additional quantification equipment such as the Agilent Tapestation, Qubit or Picogreen to ensure you’re not over- or under-loading the sequencer. The technology has come on a loooong way, and with no further spoilers, I expect this trend to continue even further in 2018.
Example: Nextera DNA Flex provides the entire workflow solution – Blood/saliva to normalized library (~3.5hrs) - Genomic DNA to normalized library (~3 hrs).
- 7 hours quicker than TruSeq DNA Nano (mechanical approach to human genomes) and;
- 2 hours quicker than Nextera XT (enzymatic approach to small genomes).
No need for quantification pre-library prep if DNA >100ng. This bead based tagmentation allows for normalization and skipping of costly Qubit/Picogreen/Tapescreen as well.
5. The demise of FFPE…?
Everyone saw this coming
Why? Because it simply provides better data. I’m glad to hear that Genomics England are seeking to replace FFPE with Fresh Frozen method instead. In fact, I feel that this news firmly cements NGS as a critical and transformational technology in our every day life – altering technology/processes AROUND the central tool in the workflow: NGS. The gold-standard here is to snap freeze liquid nitrogen directly or in Isopentane. Although I’m sure other companies/methods will look to develop more simple solutions with this news.
I think this shows the maturity of the NGS field. Whereby previously FFPE samples were the gold standard because they preserved tissue morphology for the main diagnostic technique at the time: immunohistochemistry, organisations are now centring ancillary equipment around the need for NGS. Will NGS be the primary genomic analysis tool of choice in 2018? Maybe, just maybe…
Changing FFPE to Fresh Frozen – The ‘gold standard’ is to snap freeze in liquid N2 directly or in Isopentane.
6. Larger and more complex enrichment panels
People are demanding more from NGS, and enrichment techniques are providing.
Contrary to recent announcements (eg. AmpliSeq for illumina – who saw that coming?) enrichment techniques are paving the way forwards in the panel-based sequencing space. Why? Well let’s break down the comparison:
|Content||Up to 500 kb||500 kb‒15 Mb|
|Insert size||150‒425 bp||250‒300 bp|
|Input||10 ng – intact DNA
>10ng-250ng degraded DNA
|Types of variants||Short Indels, SNV, known breakpoints||Short and long Indels, SNV, CNV, fusions|
So there are clearly a lot of advantages, and let’s be honest there are a few key disadvantages too if you work in the clinical space (eg. FFPE compatibility). But overall, enrichment techniques are likely to be adopted by labs and produced by NGS companies as the trend towards not only using NGS for diagnostic purposes, but also prognostic and recurring/routine monitoring. Furthermore, InDels, CNVs and VUSs are only ever going to become, well, less VUS’y.
So that’s it from part 1. In part 2 I aim to include some of the other mammoth topics that are as important, but unfortunately missed my top 6. Including: multi-omic approaches, single cell RNAseq, clinical WGS and longitudinal sequencing approaches. Any suggestions or feel I’ve missed something important? Comment below.