Name: Marilyn Li
Title: Director, Cancer Genetics Laboratory, Baylor College of Medicine
Marilyn Li may be a pediatrician with a focus on cancer genetics by training, but she has also become an advocate for the adoption of genomic technologies such as chromosomal microarrays and next-generation sequencing to better study, understand, and ultimately help diagnose and treat cancer patients.
As director of Baylor College of Medicine's Cancer Genetics Laboratory, she has used custom cancer-specific microarrays to study hematological malignancies to identify copy number alterations that can be used for risk stratification, therapeutic interventions, and disease follow-up. At Baylor, she has also implemented array-based testing for solid tumors. And she recently served as the Cancer Cytogenomics Microarray Consortium's president, overseeing a cross-laboratory performance study of multiple platforms.
Li has also adopted next-generation sequencing in her lab, designing targeted panels to detect multiple pathogenic mutations in the same assay. She discussed the opportunities presented by these technologies at the recent Association of Genetic Technologists, held in Las Vegas last week. BioArray News spoke with Li after her presentation. Below is an edited transcript of that interview.
One issue that you raised in your talk concerned the decision to use lower-density CGH+SNP arrays sold by Agilent Technologies versus the higher density SNP chips sold by Affymetrix and Illumina. There is a perception that higher resolution is preferable. Do you agree?
I think both platforms are good for clinical applications. Each array platform has its advantages and disadvantages. It's not that the higher the density, the better. When you make an array with higher density, you have to use more probes. Probes do not behave exactly the same. Some of them behave very well, but some of them don't. There are some polymorphic features associated with it. So when you get higher density, your array could be noisier, because the quality of some of those probes may not be as good as others. Therefore, the resolution of the array may not increase as much as you expect. There should be a balance between the number of probes and the quality of probes. In general, CGH arrays work better for copy number variations but SNP arrays provide genotype information that allows the detection of uniparental disomy and copy number neutral loss of heterozygosity, the latter is very important in cancer. CGH arrays alone will not provide genotype information. But, overall, the quality of arrays from all array vendors has improved a lot. CGH array vendors have now added SNP probes into their CGH arrays for the detection of UPD and LOH, and SNP array vendors have adopted copy number probes to cover so-called SNP deserts, the genomic regions that have few SNPs. One good thing about CGH arrays is that you can customize them, depending on your need. If you are doing cancer, you can put a higher density of probes in cancer genes. If you are going to study a specific disorder, and you know that the disease is associated with copy number change of a specific region, you can put more probes in that region. That is why at Baylor, we use an Agilent 2x400K CGH+SNP array and there are two types. One type is specific for cancer. On the constitutional side, we have another array that targets different genes associated with genetic disorders.
Who designed that array?
We designed it ourselves.
Is it the mission of Baylor to drive the adoption of new technologies? Your colleagues Art Beaudet and Jim Lupski are pediatricians, your focus is in cancer. And yet you all have spent a lot of time designing and advocating the use of arrays and other technologies.
I think for Art and Jim and me as well, it is a passion. They are pediatricians and scientists. They see genetic patients in clinic and want to make sure they can make the right diagnosis for those patients. When new technology that has the clinical diagnostic potential comes out, they want to adopt that to help their patients. And it has been the same for next-generation sequencing as well. We were the first to offer whole-exome sequencing to patients who have been through the whole ordeal many years without diagnosis. Now we have identified causal mutations for many of them. For me, it's the same thing. I was trained as a pediatric hematologist. I had seen young patients with leukemia and lymphoma. I treated them, I watched them die. It's very, very hard to see your patient die. And it's not just that you touch patients' lives. They touch your lives as well. So, cancer genetics has always been in my heart. When microarray technology came out, I immediately thought that it would be a great tool for cancer, because there are so many unknowns in cancer. I remember that when I was doing my thesis, it was about chromosomes and leukemia, and a lot of it focused on chronic myeloid leukemia. I remember writing that one day we might be able to find a treatment based on genomic aberrations, and we did. Now we know that can be done and CML patients can be cured. They don't have to die. Research is a great thing. But if we don't translate those results into patient care, it means nothing. And that's what we want to do at Baylor. That's our mission.
How do you go about translating those findings into clinical care?
If we see a technology that has potential clinical utility, we will try it. It's not very easy to do. We have to be very thorough. It starts with a literature search, identify the technology, design the test, validate the test, and that's just the R&D part. Once the test is done and ready to be sent to the lab, two technicians have to validate it with multiple samples and come up with the same results. We have to be very strict, because it is going to make a difference in patient's treatment. Like in next-generation sequencing, we were the first ones to offer it for clinical diagnosis. The panels were first run on samples with known mutations, that is, I wanted to see if the panels could detect those mutations. Once that was done, we did a series of dilutions to see at what sensitivity we could detect those mutations. Then we got 50 different formalin-fixed, paraffin-embedded samples from our tumor bank. I had no clue what they were or what mutations they had. We ran those on the panel and then used a secondary technology that was already validated in the lab to confirm those mutations. After that was done, and we have written up the standard operating procedures, we sent it to the lab for parallel validation by lab technicians. Once that was finished, the validation record was sent to our laboratory Quality & Compliance Management officers for review. Once approved, it's ready for use.
Some of the speakers here have said that CGH is well suited for hematological malignancies, but is not there yet in terms of testing solid tumors. What is your opinion?
I would say that microarray should be the first-tier test for solid tumors. The main reason is that with leukemia and lymphoma you can easily get chromosomes. If you obtain bone marrow and capture the cells it is much easier to get chromosomes to analyze compared to solid tumors. Number one, they may not grow. Number two, if you do get metaphases, they are so short and complicated that you might not be able to analyze them. But if you run microarray, it's clear. You know the copy number changes. The only thing that you are going to miss is those true balanced translocations. That's why we are working with different technologies. Maybe we will come up with a sequencing panel that will specifically target translocations.
How do you see those technologies being used now? It seems that people are using a mix of technologies to answer different questions.
I think mutations as well as copy number changes are very important in cancer, both germ-line mutation as well as somatic mutation. So right now, with current technology, array CGH and SNP arrays are the best for copy number. Next-generation sequencing is the best for mutations. But I think that chromosomal microarrays will eventually be replaced by next-generation sequencing. The copy number data is there within the next-generation sequencing data. But we haven't found a good way to extract the copy number data from the next-generation sequencing data yet. So, the main reason why next-generation sequencing is not being used for CNVs right now is the software issue. Certainly, we also need to adjust the chemistry to ensure our coverage is more even. Then the computation will be easier.
When you talk about sequencing, are you talking about whole-genome, whole-exome…
Right now, I think that targeted sequencing is appropriate for cancer because you can't wait. Whole-genome or whole-exome sequencing takes four months. There is no cancer patient who can wait that long. Targeted sequencing takes a few days, and that makes it suitable for cancer care. It has been well received in academic cancer centers, like Baylor and Johns Hopkins. In the community cancer centers, it has not been adopted as widely. They are still ordering one gene-one mutation tests, because those are in the standard guidelines and they go by the guidelines. I think this is a matter of education, because if you order KRAS, BRAF, EGHR, that's going to be more expensive than running one panel. Insurance reimbursement is another issue. Insurance companies compare one test to one test. They don't compare the data you can get from one test to another test. I also think that the organizations that make the guidelines need to see more data come out before they can recommend a new test.
But even if you have that data, how will it help inform treatment?
Some information is available already. For example, based on microarray data, we can classify renal tumors. Of course, that is diagnostic significance, not treatment significance. For treatment significance, if you run an array and find a lung cancer or breast cancer patient with EGFR amplification, the patient may benefit from an anti-EGFR targeted treatment. I think some recurrent gene amplifications could be driver mutations that would be likely therapeutic targets.
Can you provide an update on the CCMC?
We will have a meeting this year on August 5 through 7. Registration is going very well. We might run out of room. The ACMG recently published guidelines for cancer microarray analysis. I think this year we will see more people adopt microarray technology in cancer diagnosis. We have received some excellent abstracts. And you also see this year abstracts about next-generation sequencing in cancer diagnosis. The CCMC will not be just for microarrays anymore.