Name: Saskia Biskup
Title: CEO of CeGaT, Germany, since 2009; clinical director, Institute of Clinical Genetics, Stuttgart, since 2012; group leader, Hertie Institute for Clinical Brain Research, Tübingen, since 2010
Experience and Education:
Junior group leader, University of Tübingen, 2008-2010
Research fellow, Johns Hopkins Medical Institutions, 2005-2008
Research fellow, Technical University Munich, 2002-2005
PhD in genetics, University of Würzburg, 2003
MD, University of Würzburg, 1999
In 2009, after returning from a research fellowship in the United States, Saskia Biskup, along with her husband Dirk Biskup, founded CeGaT — short for Center for Genomics and Transcriptomics — in the university town of Tübingen in Germany.
The privately held company, which is profitable and had about €2 million ($2.6 million) in revenue last year, started by offering single-gene diagnostic tests for rare diseases, as well as high-throughput sequencing services for research on the SOLiD platform. In 2010, it launched its first diagnostic sequencing panel, for hereditary eye diseases, which includes more than 200 genes and runs on the SOLiD 5500xl. Since then, the firm has developed diagnostic panels for many other disease areas, including neurodegenerative diseases, epilepsy, neuromuscular diseases, hearing loss, and hereditary cancer (IS 10/26/2010).
Earlier this year, CeGaT received a single-digit-million euro investment from German healthcare technology company B. Braun, which took a 20 percent stake in the company, enabling it to invest in additional sequencing platforms and to grow faster.
Clinical Sequencing News met with Biskup at CeGaT last month to talk about the company and its work. Below is an edited version of the conversation.
How many diagnostic panels do you offer right now?
We offer more than 60 panels. Our strategy is to combine overlapping phenotypes in one panel. Take, for example, our panel for neurodegenerative diseases. We started with Parkinson's disease, but lots of patients also have dementia, so we included dementia genes, as well as other genes, for a total of 280 genes.
We restrict the analysis first to a subset of genes that are very likely to be causative. The good thing about our large panels is that we can bioinformatically extend the analysis to the other genes. So if the clinician says, 'The phenotype is Parkinson's disease, but some dementia is also associated,' then we can extend the analysis to the dementia genes, and so on. We have a much higher chance to identify the causative gene by this large panel approach.
Panel sequencing is a screening method, because next-generation sequencing is a cost-efficient, very fast method, but sequencing errors occur. We use the SOLiD 5500xl platform as upfront screening, and then when we discover a hit, a variant that could potentially be disease causing, we perform Sanger sequencing to validate it. So every medical report we send out is based on Sanger sequencing.
How many staff do you have, and in what areas?
We have around 25 employees, and we are growing. We have a head of DNA diagnostics, who is responsible for single-gene analysis of more than 300 genes. We just hired a very experienced medical doctor specialized in human genetics, who will start shortly as head of panel diagnostics. We have a head of next-generation sequencing, who is responsible for running all samples on the NGS platforms. We have R&D, which we include in our processes since we optimize enrichment sequencing all the time and change protocols rapidly. We also have a bioinformatics department, and we have marketing and sales. Within the DNA diagnostics and panel diagnostics areas, we have scientists who are specialized in certain disease phenotypes.
How did you choose the SOLiD platform for your tests?
[Before we launched], we talked with all the major companies — Roche, Illumina, and, at that point, Applied Biosystems [which merged with Invitrogen to form Life Technologies in late 2009]. As we knew that diagnostic testing would be our main focus, we wanted a platform with very high accuracy. And at that time, we were very impressed by the two-base encoding of the SOLiD sequencing-by-ligation method, so we decided to use the SOLiD system. We started with the SOLiD 3 platform. We then upgraded to the 3.5, 4, and now we work with the 5500 xl. We have two SOLiD 4 and two 5500xl. Since the 5500xl is very new, we had to keep the 4's, since our diagnostic pipeline was very well established on the 4. In the meantime, we have established our diagnostic panels on the 5500xl, and in the future, we will most probably only work with that platform.
We recently also invested in the Ion Torrent platform and will also order the Ion Proton soon. I'm convinced that a combination of two platforms is very interesting for diagnostics as you have platform-specific errors, and if you combine two technologies, you can reduce the number of variants that are potentially relevant for the disease dramatically.
Also, the SOLiD is a short-read platform — we sequence only 75 base pairs — which I think is not a limitation, especially in the coding part of the genome. But with the Ion Torrent, we can have longer reads, so you can sequence the backbone and then put the coverage on with SOLiD reads.
What do you have to do to start running diagnostic tests on the Ion Torrent?
The main focus of CeGaT was the development and the design of gene panels, which we enrich using Agilent's custom in-solution method. From exome sequencing, we know that there are large parts of the coding regions that are not covered well enough for diagnostic purposes, so we had to do our own design. Every half year, we come up with a new design and improve the enrichment kit by including new baits, taking out other baits, balancing kits.
We want to use the experience we already have with our Agilent in-solution custom enrichment kits for the Ion Torrent. Amplicons are maybe not the best options for our large panel approach. I think they are very interesting if you have smaller sets of genes, but we want to sequence between 100 and 500 genes, and we want to use our enrichment design on the Ion Torrent. This is not yet established, but we are very confident that this will soon be transferred.
We already have a project where we sequenced tumor and normal tissues, and we used Agilent in-solution enrichment on SOLiD and TargetSeq on Ion Torrent. We combined both datasets and found that it has significant advantages over using only one platform.
Is tumor sequencing a possible application of this two-platform approach?
There are two companies located here in Tübingen, Immatics and CureVac, that have a very interesting approach. The hypothesis is that a tumor has specific mutations that are presented to the immune system, and by sequencing, you first identify these tumor-specific mutations, and then you synthesize peptides that can be used for vaccination. The idea is that the immune system only kills the tumor cells and does not affect the normal tissue. Sequencing is one part, and you can imagine that false positives or false negatives would be of dramatic consequence here.
So we need to find a way to sequence very fast and identify these tumor-specific mutations, and we have to be very sure, and we don’t want to do a lot of Sanger sequencing to get to the list of tumor-specific variants. This is why I think the combination of the two platforms is a very interesting approach. We just did this experiment, and we are very confident that it will be the way to go for us in the near future, this combination of Ion Torrent or Ion Proton and SOLiD sequencing.
Will panel sequencing move to exome or whole-genome sequencing in the future?
I think the panels will still have a place in diagnostic sequencing for quite a while. In Germany, we have the gene diagnostics law, which says that if you have a specific medical question, you should limit your investigation to a set of genes. You could use exome sequencing and then bioinformatically restrict the analysis to the list of genes that you are interested in, but it has significant disadvantages, and I think it will also have those in the future.
The first is that the enrichment of these genes of interest is not good enough — the coverage is not high enough for the sequence to have diagnostic value. The cost is also still a dramatic difference. For exome sequencing, we can run, let's say, six to 10 samples in one run, and with panel sequencing, you can run up to 50 samples in one run. So it's not only that we get much higher coverage on the genes of interest, but we can also run many more samples. Cost is a very important factor — we try to reduce the cost of our panels, and this is the only way of doing that. I think for this reason, it makes sense for at least the next several years to have panels as an option.
Nevertheless, there are certain diseases, for example mitochondrial diseases, where you have more than 1,000 nuclear-encoded genes, where you would definitely use exome sequencing in the first place. Or other very rare diseases, where you do not know any causative gene.
We already apply exome sequencing in a diagnostic setting. It's also possible that we proceed stepwise. For example, for some diseases, a gene is sequenced first. If negative, we do a panel, and if the genes on the panel are negative, we do exome sequencing.
How much do you charge for a gene panel?
Our panels all have the same pricing. We proceed as follows: as a first step, we do next-generation sequencing with very high coverage, and the cost for this is €3,000, no matter what panel we sequence. In the next step, we validate potentially disease-causing variants, and the costs are €100 per variant. So in an ideal scenario, you have next-generation sequencing, plus two variants, plus a medical report, which is also €100, so €3,300 in total. It also happens that we ask for segregation analysis, so we analyze parental DNA, and then we also have €100 per variant and the costs become slightly higher.
Are the diagnostic panels you currently offer being reimbursed by health insurance?
No, not yet, but it is now becoming clear that there is no way around the panels. I think everyone who is doing genetic testing knows that the panels have a huge advantage with respect to certain diseases, and I think it's only a matter of time until there will be reimbursement for panels.
If you take epilepsies as an example, if you sequence one large gene, including deletion analysis, the costs that the German system reimburses would already enable you to do whole-exome sequencing.
How are you reducing the cost of your panels further?
At the moment, we are at a point where we can run 48 patients in one SOLiD run. With the [Agilent] Bravo automation, you can enrich up to 96 patients, so our main focus for reducing cost in the long term is by automating the whole procedure, so samples come in, DNA gets extracted, and then 96 samples are enriched at the same time and sequenced in two flow chips on the SOLiD. Then we would also have the downstream bioinformatic analysis completely automated to the point where we come to the list of variants, and also to the list of unrepresented regions.
We want to do panel sequencing with the highest possible quality, which also means that we have to resequence some regions that we cannot enrich. The unrepresented regions will automatically go into resequencing, and by this, we can be very confident that we did not miss anything.
At the end, an individual who is very familiar with the disease analyzes the data and writes the report.
In terms of the clinical interpretation of variants, how much is already automated, or can be in the future?
I think there is a huge gap between the drop in sequencing costs, which is very rapid, and on the other hand the costs and value of the analysis. I think it's important that someone who is really familiar with the disease, who knows the genes and the variants, looks at the data, and I don't see a computer doing this at the moment.
A computer can do a lot; we do lots of comparisons with external and internal databases, human mutation databases, [and so on], but still, it's very important that someone sits there who looks at the data and bridges the gap to the clinician. We have a patient with symptoms, and sometimes that doesn't fit to what the SNP list is telling you, so I think it's very important that there is a conversation between the clinician and the human geneticist, also, and there has to be a person who is responsible for this part, and this is still expensive.
Do you believe that you might eventually outsource the sequencing part and only focus on the clinical interpretation?
No. We tested this. We did some collaborations and we found out that our standards in the laboratory, and the way we automated the procedures, are very important to obtain confident and reproducible results. So for us, it's much easier if we can validate our own data, and then write a report, compared to either sending out NGS data, where someone else is doing the interpretation, or getting data where we have to do the interpretation, because we have no idea about the quality of the sequence. We do not know what is missing, how many regions are underrepresented. In Germany, also, if you offer testing, you need to do this all in your own laboratory.
How are next-gen sequencing-based diagnostic tests like yours regulated in Germany?
Not at all yet. There were no panels available when we started, and we established everything from designing the panel to the final report. I think there are now groups doing diagnostic next-generation sequencing in Germany that try to develop a list of things that have to be fulfilled before you can offer this as a diagnostic test, but we are still in the process of coming to a point where we can say, 'This is now comparable from one laboratory to the next,' since everyone uses a different platform, different enrichment strategies, so it's very hard to get a feeling for the quality of the data.
What requirements are there in Germany to offer any diagnostic test? Is there an equivalent to CLIA [Clinical Laboratory Improvement Amendments] in the US?
Yes. We have European accreditation and ISO certification [DIN ISO EN 15189:2007; DIN stands for Deutsche Industrie Norm; ISO for International Standards Organization; EN for European norm] that we need to have as a human genetics diagnostic lab. We have been certified since February of last year.
Several vendors now want to submit their sequencing platforms to the US Food and Drug Administration for 510(k) clearance. How important would that be for a provider like CeGaT?
It would not be important for us since we are always validating our results by the gold standard, Sanger sequencing, or quantitative PCR. What's important for us is that the sequencing and enrichment companies are aware of the fact that we are working with patient data, so they get out the highest possible quality products. Sometimes we also have long wait times for receiving kits, for example, and there is, I'm sure, room for improvement, but for our reports, certification of the instruments is not necessary.
How do you deal with variants of unknown significance, or with novel variants?
If we think that these variants might have some clinical significance, we suggest doing segregation analysis, if possible. This is for variants where there is nothing known in the literature. For our panels, we collaborate with clinical centers for the respective diseases, and we have the opportunity to include novel variants in functional assays. It can take a long time until you find out whether they are really the cause of the disease or not, but it's important to be done.
If we are not sure, we report it and say we are not sure at this point, but we do not report variants where it's most likely that they have no significance. We write in the report that if the variant that we suggested is excluded as a cause, we can always go back in the data and validate other potentially interesting variants. But especially for exome sequencing, there are so many variants, it's not possible to validate them all in the first place.
What kinds of results do you notify patients about?
For every test we do, we talk with the patient or the referring clinician and explain what can happen. One thing that can happen is that you find something by accident. So you are looking for a muscular gene, and by chance you find a tumor gene. The patient has the option to say if he or she wants to be notified. They usually decide to be notified only if this has a therapeutic or other consequence.
Patients know that DNA is anonymized in a scientific study, and then there will be no notification about the results. I think it's also important to make that clear to the patient.
What new panels are you planning to develop?
We were approached by several centers that think the panel strategy is a good way for their patients as well. We just developed a deafness panel that contains more than 100 genes; we closely collaborate with the Comprehensive Cochlear Implant Center Tübingen on that.
We are also now in the process of developing a pharmacogenomics panel that includes almost 500 genes. This will be a research panel that we envision pharmaceutical companies might be interested in using. But we are also developing smaller panels for particular clinical implications, where there is a specific disease in which some medication is known to be more efficient or less efficient.
We also just developed a tumor panel for germline mutations, including tumor syndromes, and one for somatic mutations. We are used to dealing with tumor samples through the research arm of our company, where we do lots of tumor sequencing. We sequence fresh-frozen but also fixed samples, and we know of the problems associated with contamination of tumor tissue with normal tissue, the problems of coverage, and interpretation and validation. What I think is important for tumor diagnostics is combining next-generation sequencing with other available methods. We still do [fluorescent in situ hybridization], for example, and array [comparative genomic hybridization] and NGS sequencing with Sanger confirmation, so we offer the whole package.
What diagnostic area do you think is going to be most important for your business in the future?
Worldwide, I would say that tumor diagnostics is going to be a huge market. For my own research, I think that genetic testing in neurodegenerative diseases will be a major part in the future. It's not a big part at the moment since no therapeutics are available, so no one wants to know if he or she is going to develop dementia, but I think once medication is available, one would want to know if there is a risk variant that predisposes to the disease, and this would offer the possibility to give this medication much earlier. For neurodegenerative diseases, it could mean that you would need to start medication maybe 20, 30 years before disease outbreak.
Pharmacogenomics, clearly, is going to have a huge part. You don't have to be sick to be in need of pharmacogenomic panels. It could be just simple medication for migraine or whatever, and you want to know the dose and which drug you want to take in order to avoid side effects.
And rare diseases in general. If you take all rare diseases together, there are so many people affected worldwide. By our new methods, we can now offer every testing [for every individual], and I think that's a major advantage.
A video interview with Saskia Biskup is available on GenomeWeb's video channel.