Name: Leonardo Meza-Zepeda
Title: Head, University of Oslo Microarray Core
Professional background: 2003-present, senior scientist, University of Oslo, Norway; 2008-present, head, University of Oslo Microarray Core Facility, Norwegian Radium Hospital; 2006-2007, senior scientist, Norwegian Radium Hospital; 2001, senior scientist and consultant, Complete Genomics AS, Oslo
Education: 2000 — PhD, cancer genomics, University of Oslo, Norway; 1996 — guest researcher, plant molecular biology, Swedish University of Agricultural Sciences, Uppsala, Sweden; 1995 — BS, biochemistry, Universidad Austral de Chile, Valdivia, Chile
With nearly a decade's worth of experience with microarrays, Leonardo Meza-Zepeda has seen the technology evolve from bacterial artificial chromosome arrays to high-resolution oligonucleotide arrays and, in recent years, pairing array-based projects with next-generation sequencing.
As head of the University of Oslo's microarray core facility at the Norwegian Radium Hospital, Meza-Zepeda offers researchers in Norway the ability to run studies on Agilent Technologies, Affymetrix, Illumina, and Roche NimbleGen platforms. He and his colleagues recently took advantage of their access to the four largest array vendors' tools to compare each one's ability for mapping DNA copy number aberrations in bone tumors.
In the study, published online in BMC Research Notes this week, the authors tested DNA from two different cytogenetically well-characterized bone sarcoma cell lines, representing a simple and a complex karyotype, respectively, in duplicate on four high-resolution microarray platforms: Affy's Genome-Wide Human SNP Array 6.0, Agilent's Human Genome CGH 244A, Illumina's HumanExon510s-Duo, and NimbleGen's HG18 CGH 385K WG Tiling v1.0. The data was analyzed using platform-specific analysis software as well as a platform-independent analysis algorithm, and DNA copy number was measured at six specific chromosomes or chromosomal regions and compared with the expected ratio based on available cytogenetic information.
According to Meza-Zepeda's findings, the major vendors' platforms performed well in terms of reproducibility and were able to delimit and score small amplifications and deletions at similar resolution, but Agilent's microarrays showed better linearity and dynamic range. Additionally, while the platform-specific analysis software provided with each platform identified in general correct copy numbers, when using a platform-independent analysis algorithm, correct copy numbers were determined mainly for Agilent and Affy microarrays.
Meza-Zepeda discussed the outcome of the evaluation with BioArray News this week. Below is an edited transcript of that interview.
What is your background and how did you wind up at the Norwegian Radium Hospital?
I studied biochemistry in Chile. When I reached the level of my master's degree, I was in Sweden at the University of Agricultural Sciences. After that, I decided to move to Norway to do a PhD, moving into the field of cancer research. I did my PhD, a postdoc, and now I am running a core facility for regional health authorities. The facility is part of the national Norwegian Microarray Consortium. We have a microarray core facility and now we also have next-generation sequencing, which we have taken as a further addition to the array services we have provided for over 10 years.
We have been studying soft tissue and bone tumors for some time. Part of my PhD work was identifying genes that were correlated with bone tumors. At the end of my PhD, I attended the University of California, San Francisco, where I learned to make bacterial artificial chromosome arrays. Then I took the technology into Norway and developed BAC arrays here for studying tumors and constitutional aberrations.
When was that?
That was in 2001. When I was at UCSF we developed a focus array for chromosome 1. The year after [that] we got a clone set from Nigel Carter [group leader of molecular cytogenetics at the Wellcome Trust Sanger Institute] and expanded the arrays to 1 megabase coverage. We used those arrays for a very long time to profile a number of tumors, osteosarcomas, leiomyosarcomas, lymphomas, and others. More recently, we shifted to newer technologies like oligonucleotide arrays. Within EuroBoNet [a European network that promotes bone cancer research], we are providing services for DNA copy number variation studies. We decided at that point to evaluate different platforms to see which ones perform better in highly complex tumors like bone tumors. EuroBoNet was started in 2006 and the evaluation project was carried out at the end of 2008 and the beginning of 2009.
Why was a comparison necessary?
The problem is that bone tumors are highly complex. They have an extreme amount of aberrations and sub-populations within the tumors. We wanted to test different platforms with regards to reproducibility and the ability to identify subpopulations within a complex mixture of cells. In general, where you don't have pure population of cells, it is important to have platforms that can identify the, say, 30 percent of cells that have a certain aberration.
How did you select the arrays for your comparison?
We selected the arrays according to density of probes and coverage. We took what was available [at the time]: the Affymetrix SNP 6.0, the Agilent 244K, and the [Roche] NimbleGen and Illumina arrays. We offer services for all of those platforms. I should add that some of those platforms, like the NimbleGen and Illumina and Agilent, have newer versions at the moment.
[ pagebreak ]
How did you conduct your evaluation? What did you measure?
We had two different cell lines, one with less complex separation but different levels of copy number changes in different chromosomes. We looked first at reproducibility between different platforms. We also looked at signal response to copy number changes at different chromosomes and how linear that response was. We also looked at how easy it was to detect aberrations that we knew existed with the vendors' proprietary software packages. Then for our analyses we also used Nexus from BioDiscovery as third-party software because it [makes] it easy to import different platform data from Agilent, Affy, Illumina, and NimbleGen and compare.
Can you describe your findings?
All the platforms offered more or less equal reproducibility. We saw the main difference was linearity, the response to increases in aberrations in copy number. What we see is that Agilent saw the highest dynamic range as well as linearity. Affy also showed a reasonable performance — it was the second best one with regards to dynamic range. And then there were Illumina and NimbleGen. Some of them had deviations larger at the extremes, when you had four copies of very low copy number.
Were there reasons, in your opinion, that certain platforms performed better than others?
It would be difficult to speculate as to why. Agilent offered larger oligos, and is more gene centered compared to other platforms. We did not try the 2.1M platform that NimbleGen has now. We saw that, in general, oligo arrays are very noisy. This noise can be compensated by an increased number of probes and using mathematical smoothing using a moving window. There is one figure in the paper that shows the distributions of the different copy number states measured by arrays. The separation of the different copy number states was improved considerably by applying a Gaussian smoothing with a window size of 50 kb. So the amount of probes will not only help to identify smaller regions but also to better separate the distributions of log2 ratios when mathematical algorithms like smoothing are applied
Does this mean that you will recommend Agilent and Affy for these kinds of studies?
Our conclusion is that Agilent and Affy seem to have more robust platforms for measuring DNA copy number changes. Depending on application, you can use SNP arrays from Affy, but if you are not interested in SNP data, Agilent is also good alternative.
Do you prefer the SNP platforms or the CGH platforms?
The benefit we see with tumors is that SNP information can give you an opportunity to identify copy number neutral loss-of-heterozygosity regions where one of the paternal alleles is deleted and restored with a defective allele. That is an advantage when working with cancer.
You are hosted by the Norwegian Radium Hospital. Have you translated any of your findings to the clinic?
We have identified candidate genes, genes that have some correlation with clinical features. We have identified a gene correlated with osteosarcomas called LSAMP that is frequently deleted or methylated. Some of that information can be used to create biomarkers to select patient groups that benefit from more aggressive treatment. Right now, this work is at the level of basic research, but it may have in the long term the opportunity to identify genes that are biomarkers and could be used to design new drugs for targeted therapy.
How is sequencing being adopted?
I would say that sequencing serves similar researchers, but with questions that were not able to be answered by arrays. We are interested in my research group in studying fusion genes, things that were very difficult to identify with arrays. Now we are sequencing the transcriptome to identify fusion genes in osteosarcoma. EuroBoNet is ending now, but as an extension of the consortium we are looking at sequencing the tumors to identify special translocations that can be used as markers to follow up on disease. In other cancers, they have identified a special rearrangement by sequencing, and you can design specific PCR assays to monitor tumor DNA in blood or in plasma. These can then be correlated with response to treatment or disease-free state, a personalized biomarker.
And how will this impact your use of arrays?
Arrays will have [their] niche if they are relatively inexpensive, and it is possible to do hundreds of samples in a small lab to look at CNVs or the transcriptome. You can also use arrays for validation. You can design custom arrays to validate hundreds or thousands of candidate fusion transcripts, or, if you like, to screen a large panel of samples. I think it is a very complementary technology. We have stopped producing BAC arrays though. We have moved everything to oligo arrays.