#AGBTPH – The Great Debate (Panels vs. Exomes vs Genomes)

  • Richard Gibbs, Baylor College of Medicine
  • Heidi Rehm, Harvard Medical School, Partners Healthcare
  • Steven Kingsmore, Rady Children’s Hospital

Heidi Rehm is up first.

  • No one test fits all.  Should be influenced by phenotype, insurance, gene of interest, etc. Some genes are incredibly hard to work with (e.g.. hearing loss genes) so many genome and exome approaches, so panels would be better.  30% of hearing loss issues are CNV, which are very difficult to detect by exome.  If diagnosis is really tough, and there is no panel that caters to the symptoms, then go for genome.
  • We used to operate with labs that focused on single genes – but now our labs are broad, but we’ve lost all of the experts on single genes.  knowledge management is very hard in this environment.  Sometimes genomes miss things that panels catch, and vice versa.   At end of day, it’s all context specific.  If you have to factor in insurance coverage – and insurance companies really aren’t interested in secondary findings.  etc.
  • Address all of the other complexities: SVs, CNVs, etc. etc.  But these are expensive, but physicians will still go with the cheapest labs.

Richard Gibbs is next

  • The two points of view here are :
    • whole genomes are better
    • no, it’s more nuanced, it depends on the context
  • Most of us are probably in the second camp, if we were pressed.
  • The quality of the genomes, exomes and panels is still a huge factor.   Genomes DO miss things.
  • Exomes are getting better – we’re now up to about 84% coverage of every base in every gene in the coding genome. All genes are there, and most are only missing a small number of bases.
  • Genomes cover everything, but calling is a big deal.
  • CNV coverage is poor on Exomes, but good on genomes and panels.
  • Cost is linear between panels, exomes and genomes.  Informativeness relatively linear as well, but you can explode cost if you want to do more on a genome.
  • If you have a choice between trio of exomes vs genomes, you need to consider that too.
  • The vendor is selling instruments that have distorted prices for exomes, which could be as cheap as panels, if they wanted.

Steven Kingsmore:

  • Going to be provocative, and push hard. [devil’s advocate]
  • The Genome IS the ultimate test, and it’s incredibly informative.  Everyone needs a genome!
  • Genomes
    • Doesn’t miss all exomes of all genes
    • You don’t miss all intronic variants you’d like to see.
    • You can look at all of the reads to see what you’re missing in the gene of interest.  can’t really look at that in panels and exomes.
  • Problem is that they’re expensive.  At least 4 times more than an exome.  (and that’s 4 times more than a panel) Payers don’t pay us much.  How much money do you want to lose.  People don’t mine the genome either – they’re doing genomes, but only get back what they’re looking for.
  • Special cases, though: Cancer needs to go REALLY deep – 30x isn’t enough for somatic mutations.  Exomes will win there.
  • Hybrid Genomes – the next great things: matching long read sequencing with short read sequencing. Rich mining of indels/SVs.  SO EXCITING!
  • Panel < Exome < Geomes << Rich/Hybrid Genome (EVERYTHING ON STEROIDS….  which costs even more.)

Moderator: Why are hybrids a good thing? why is it so much better?  We can already get SVs, but what we can’t see, why would be expect to see it then?

RG: Just on the cusp of giving you all the rich information that we want.  If cost wasn’t a limit, we’d probably come up with a $40,000 bit for each genome.  Yes, we want rockets to mars, but the cost is much more nuanced now.

SK: Everyone is spending the same amount on a battery of tests, but genomes give us that anyhow.

RG: But we’re still missing stuff because of the technology.  You’ll still miss critical expansion regions anyhow.

Q: We’re so far beyond where the physicians are, but ultimately, we’re dealing with uninterpretable content in the genome.  Why is this really relevant?  The real question: there’s value in all of these, but if we want this to translate to the clinic, how we convince them any of this is useful?

HR: The strategy for reporting is very different between panel, exome, genome.  Panels, require an investigation of every variant in gene.  Even if you have VUS, you can follow up. In genome exome, you can not interpret every variant, so you filter.  Sometimes those filters don’t do a good job.  So, we miss things not because they’re technically missed, but because the clinicians and bioinformatics are failing to do what they should.

First lawsuit for misinterpretation of variant – the lab reported the variant, and requested parental testing in order to augment to pathogenic.  The doctor didn’t pass any of that on to the family, and the family found out years later.    The physicians have to be brought into the conversation.

SK: Have to respond to HR.  Panels tend to overdiagnose.  You have 10 genes, and darn it, you’re going to make a diagnosis.  You tend to want to call something.  Under diagnosis is the other end, because you miss other genes. Panels are cheap, and that’s why we do it, genomes will eventually be there.  Respond to questions on Physicians: moving to hybrid genomes will help, because then physicians just have to know one word: Hybrid-genome.   They don’t need to know protocols and panels… Patients shouldn’t have to bounce through specialist till the right test is run.

RG: Is it true that panels are over diagnosed?  It is a persistent problem.

HR: See just as many over diagnoses from genomes and exomes – it’s not unique to panels.

Q: (Fawzan) Point 1: litigation, Point 2: variant is there, but we fail to see it.  That combination is terrible, so we’re amplifying the odds of being sued!  We aren’t discussing that enough.

HR: I don’t think any of us are liable for doing the tests that are requested or doing their best for the patient.  Whoever it is, if you’re doing your best, follow your protocol, then you’re not open to liability.  If we diverge from protocol, or don’t validate, that’s when we become liable.  There’s not always a right or wrong.  We should be doing better, but that’s not unique to medicine.

Q (Follow up): If you’re in a court room, that 30% -40% difference in diagnosis rate may play differently.

HR: That’s why we do so many validation tests – The bioinformatics is maturing, and early versions didn’t do a great job – but it’s improving all the time.  It’s easy to say what should be done, but technically doing the pipeline is much more complex, data is more complex, nomenclature is not perfect.  Filters are very very hard to do, and pipelines need to be validated extensively.  it’s a challenge.

Q (Follow up):  When we miss something it’s usually because the filters are wrong – and with genomes this is just again opening up to liability.

SK: Your argument is silly.

Q (Follow up) : just being provocative!

SK: [reductio ad absurdum]  Patient goes to doctor with a headache…

Q: It’s contextual, we all agree.  In the practice setting, where diagnostics are being ordered.  The constraints on the doctors are even more tight.  How do you think about getting any of these tests (even panels!) accepted outside the academic environment.

RG: Kingsmore promotes genetic literacy by not burying people – applying filters that make the data less complex for physicians. We’ve even put filters in place to mimic that.

HR: Panel done on exome or genome backbone is good – That is a good transition.  Virtual panels are constructed in a way that appeals to the physician to mirror the standard of care for an off-the shelf test is a good intermediate – and allows physicians to return to that data and unmask the next set of genes.  iteratively go forth to reduce the cost, but not change complexity for the physician.

(@notSoJunkDNA): This debate ignored cancer – for another day.  We see resources like Exac, which help aggregate data.  Thus Whole Genome is the only investment in the future – which should be factored into the cost.

HR: Yes, by building that resource, we are making huge impact. Exac is single most useful resource of the last 20 years.  However, we can’t put that expense on our patients.  We can’t be shortlisted, we still need to care for patients.

RG: want to use a different argument to disagree with SK.  Whole genomes are not good enough yet.  Lets not burn all our dollars now before the genomes are great quality, lets get to the great genome, then do it.

SK:  Yes, there are flaws with genomes, but the point is well made: in an ideal world we should be getting genomes on everyone, and put them aggregated into the public domain to allow us to tackle major issues.  Really like that idea, and willing to help subsidize the incremental cost of that. The idea that you get one report, and can return to the genome and reanalyze them makes a huge impact.

M: We can generate them, but we can’t analyze them.  It’s not a great genome if you can’t analyze it.

RG: Eventually  someone is going to go back and use better methods to reanalyze what we’ve done.

Q: Hard studies are tough, but worth doing.  Q to RG: if you had a family member, would you really do 30x WGS, with existing technology  Wouldn’t you do Germline/Somatic?

RG: It’s different between cancer and mendelian.  It’s also different from family, from discovery from managing a health care portfolio.  Hypothetical emotional questions have be separate from the data questions.

HR: we have limited resources to do analysis anyhow.

RG: Have to ask what’s advancing the research agenda.

SK: we do have a lack of objective evidence. Most of our community is doing panels like this one, where we see who argues the best.  Analysis used to be an art – parameterization takes a huge amount of work and investment.  What we’re doing with sequencing is the same, and when the hard data shows up, we’ll convince the payers.

Q:  What price does it shift over from WES to WGS?

SK: I don’t think we’ve done clinical utility studies – it’s the missing piece.  There are studies starting to look at this.  There are way more studies needed for cost effectiveness.

HR: it’s not just prices, we’re losing something with the broader scope.  It’s a tradeoff that has to be examined, outside of cost.

M: RG, If you could put a certain test on a certain machine.  If you could run your exome on a different machine, are policies limiting the best highest quality health care?

RG: conflicting answer:  SK is right, we haven’t got all the studies done.  Even if it’s free, we don’t know how it would work.  However there are medical tests that have a price inflection that is convincing.  <$100 exomes, we would open up many new avenues.

HR: Need to look at numbers, what do the old vs. new numbers say about the changing tests.

M: We could drive down costs if we could use exomes on different machines…. that’s major issue.

Q: How do you think consumer genomics is going to change our field?   (Like GMO?)

SK: Great point – while we argue over which is best, there is an impending nightmare where traditional medicine becomes eclipsed.  Like Herbal medicine, chiropractors… there are things that medical fields didn’t embrace, but public wants it.  Patients will start wandering into their physicians with results  and looking for information.

Q (Follow up): Especially if quality they bring in is terrible.

HR: Terrible backlash to GMO because there was no public debate, and that lack of debate had huge negative impact.    It’s really important that we have those discussions.

Q: Moving target scenarios: cancer mutations, antibiotic resistance.  How do we balance coverage and cost – what’s best approach in that setting?

SK: Cancer, you want both exome and genome, match germline, tumour….. [everything!]

M: Out of time!  What should reimbursement do?

HR: It should be paying, all the hype has missed the fact that this is incredibly helpful for those who get diagnosis.  It’s on the community to do a better job to communicate and share resources so we can decide which tests are the right tests.

RG: Swayed that the top down tests are actually be good.  We shouldn’t be comparing exome vs panel.. Genomes aren’t there, but we should be looking them in the future.

SK: We’re privileged to be living in the day of NGS, and we should be looking at enabling all of the options for clinicians!

Leave a Reply

Your email address will not be published. Required fields are marked *