Radboud University Nijmegen Medical Center
Name: Joris Veltman
Title: Head of microarray facility, Radboud University Nijmegen Medical Center
Professional Background: 2000-present, head of microarray facility, Radboud University Nijmegen Medical Center
Education: 1999 — PhD, Maastricht University, The Netherlands.
If microarrays are being used by clinicians en masse in any one area, it is in cytogenetics. While the slow but steady process of seeking US Food and Drug Administration clearance for various in vitro diagnostics for drug metabolism profiling or cancer recurrence probability continues unabated, actors like Signature Genomic Laboratories, Baylor College of Medicine, and CombiMatrix Molecular Diagnostics in a matter of years have turned clinical array usage into something routine for patients that lack a conclusive diagnosis of genetic abnormalities.
Last week, Affymetrix became the latest player to exploit that trend when it announced that it is participating in a European cytogenetic research initiative that will use its SNP genotyping arrays to identify the specific causative mutations in mentally retarded children.
The initiative is a collaborative effort between Affy, Radboud University Nijmegen in the Netherlands, the University of Tuebingen in Germany, and the National Health Service Regional Genetics Laboratory in Birmingham, UK.
Rather than design its own arrays or settle on a commercial cytogenetic microarray platform, the initiative has decided that Affy’s Mapping 500K arrays are the best tools available for clinical cytogenetic studies. To learn more about the use of Affy SNP chips in a diagnostic setting, BioArray News spoke with Joris Veltman, head of the microarray facility at Radboud University Nijmegen Medical Center, about the initiative this week.
Which SNP arrays are you using? Are you using the 1M SNP 6.0 Array or something smaller? [[What was the criteria for] your choice?
We are currently using the Mapping 500K SNP arrays for our diagnostic profiling in mental retardation. The reasons for using this platform and not the recently launched 1M SNP 6.0 array are that we have most experience with the 500K platform, which allows us to better interpret the observed copy number changes. We know which regions in the genome represent normal variation and which ones are likely to be linked to mental retardation, and when you go to a new platform, new and likely even smaller regions will be detected of unknown clinical significance.
Secondly, the currently available data-analysis tools are mostly oriented towards the 500K SNP arrays, not yet ready for the 6.0 array. It’s a matter of timing. So, in fact, for diagnostic applications we will take it a little more slowly, whereas for research projects we will more quickly adapt to the novel higher-resolution platforms.
I am a bit confused here because the Affymetrix press release says it’s a research collaboration but you are talking about diagnostics.
As the press release points out, this is a research collaboration. The papers we have published show we are already applying genomic microarrays in a diagnostics setting apart from this … so in that way there is kind of an overlap. The research part is finding out: what is the best platform, how is this working out, and what are the best methods to use? But as soon as this is working for us we can use it diagnostically, and we do this in our lab.
Are you following up with something like fluorescent in situ hybridization? What are you using to confirm your results?
We always confirm our results. Karyotyping is not possible because the aberrations are too small, FISH is something that we do use a lot, but actually for most applications we are using quantitative PCR assays. That is a technology called multiplex ligation-dependent probe amplification. The problem with FISH is that it is very hard to see duplications, for example, if they are occurring in tandem on the genome.
Also the problem with FISH technology is that you need to culture your cells in order to do that and the advantage of microarrays is that you can simply isolate DNA from a drop of blood and do an array and if you want to validate you can also use the same drop of blood. So FISH is not what we use, quantitative PCR is what we use for validation.
It seems like it would be a big challenge to use something like 500K and follow it up with QPCR. Why are you so confident in using a tool where there could be a higher level of possibility for making a mistake, in terms of the amount of data you have to look at?
It is clear that we are operating here at the forefront of genetic diagnostics. We have always been one of the first to do this diagnostically. The disadvantage is that it is not easy to do this at a very high throughput with thousands of patients and randomly, because in the beginning [we] really have to think about what we are seeing — so we need more time for discussion.
At the same time we shouldn’t make it too difficult. If you look at the DNA of the patients nowadays with our SNP arrays, we find approximately 20 regions of copy number variation. Eighteen of these regions we know already occur in unaffected individuals, so we only have one or two remaining, so it is not something where we spend months on every patient. We have published that the diagnostic yield is doubling in these patients by doing a genome-wide analysis, so it is taking more time, but the result is that you really find a genetic cause in double the amount of patients.
Nowadays in cytogenetics you typically pick up changes in 10 percent of patients. We pick up aberrations that cause disease in 20 percent of the patients we see. We still have many questions related to the remaining 80 percent, but at least we can give a good diagnosis in the additional cases.
You are using the 500K instead of the 1-million-SNP product. Do you see any possibility that you will upgrade to using the SNP 6.0 Array?
At this stage we know a lot about the 500K, so we can use it diagnostically without too much hassle. If we would now make a change to the 1 million SNP array it would take time to get the same amount of throughput and knowledge, so you really need to build up experience. At this point we will not do that.
The market for diagnostic arrays is developing very rapidly. I know Affymetrix is also developing other kinds of diagnostic arrays. So perhaps in a year or so we will look at this again, but for the time being we will stay with the 500K.
In terms of working together with the other groups, are they also using this technology for diagnostic purposes?
Well, I am not a clinician, but for this collaboration, clinicians in our department are involved in seeing the patients, in describing the patients — because often apart from the mental retardation there are various dysmorphisms or other things going on. Actually that is the sort of thing that has been shown that, if you have a large chromosomal aberration, then that usually involves many genes, so there can be other clinical abnormalities involved. And that’s what we see with the aberrations that we pick up with our microarrays. Often they involve many genes, and one gene may cause mental retardation but the other gene may cause facial changes or a heart malformation or something like that.
How exactly will you work with the other researchers in the initiative? Will you just be sharing samples or working on other projects?
We cooperate in many ways. We have indeed exchanged samples for validation purposes, but most important, I would say, is our collaboration on the interpretation of the results. This takes place at the clinical site, where we are standardizing the clinical description of our patients, as well as at the bioinformatics site. We are evaluating different software packages and determining standards for calling copy number changes, as well as distinguishing benign from mental-retardation-causing copy number changes.
What about informatics challenges? How are things coming along with that for people who are doing anything with arrays and cytogenetics?
It is clear that in the old days before microarrays people were doing karyotyping in these patients and as soon as they would see something like a change in a genetic band, then they would look at the parents. If the parents were not affected, then you would say, ‘OK, this is a de novo change present in the patient, this must be the cause of the disease.’ And often these aberrations were millions of base pairs long and it’s not something that happens in the normal population.
Now we are entering a different phase because you see all these copy number variations but you also see them in normal individuals. So interpretation of what is a disease-causing variation versus, let’s say, a benign variation, not resulting in mental retardation, is a big challenge. One thing that we do [that is] very similar to the conventional karyotyping is indeed looking at the parental DNA and that is helping us a great deal to pick up changes [that] are unique to the patient. But that is not all, because it could be even that you have a new change occurring [that] has actually nothing to do with the disease.
This is very challenging and it is not something that you can simply solve with a software package. We simply have to do a lot of analyses and get more data involved. So we are looking at these regions a lot. We are trying to find the characteristics that determine whether copy number variation at that region results in mental retardation. But that makes it very challenging and that’s why you have to be very careful with the diagnostic interpretation.
Why not use something that is already on the market for diagnostics, like services or chips from Signature Genomic Labs or CombiMatrix Molecular Diagnostics?
Well, there is a big difference between Europe and America in this field, and it has to do mainly with the fact that in Europe diagnostics is done within the academic setting, whereas in America what you often see is that a pediatrician sees a patient with mental retardation and then he actually has the cytogenetics done in a company. The thing is that in America you have a lot of companies that offer microarrays for cytogenetics, but these microarrays have very low resolution and they are only targeting regions of known clinical significance.
This is very good and I don’t object to that, but the thing is that the majority of patients that we see have aberrations that cause disease that are in completely other regions in the genome. We think that there may be thousands of genes that may cause mental retardation affected by copy number change. So the problem is that you really have to look at the whole genome with a very high resolution and then your diagnostic yield will be much higher than if you use a microarray from a commercial company.