Laboratories that offer array-based tests for genetic abnormalities are generally satisfied with the quality of the technology platforms available, but are sometimes challenged by the interpretation of the ambiguous data that arrays can provide.
Because of this, companies targeting the clinical cytogenetics market are investing more resources in developing software tools to support their clients as they seek to determine whether variation picked up by arrays is benign or pathogenic.
According to both test providers and array vendors interviewed by BioArray News during a workshop hosted by the International Standard Cytogenomic Array Consortium in Bethesda, Md., last week, standardizing the way in which array results are interpreted is the major task facing the space. Users of these platforms agreed and noted that they are less concerned with the quality of arrays in the marketplace than they are by the task of interpreting array results.
"The different vendors sell different systems, but, overall, array quality is pretty reasonable," said Joris Vermeesch, a professor in the Center for Human Genetics at Katholieke Universiteit Leuven, in Belgium. "The major issue is interpretation. You could take the same patient to two different labs and get two different findings based on that counselor's interpretation of the findings. So we need to introduce some guidelines about interpretation."
Nicholas Haan, CEO of Cambridge, UK-based vendor BlueGnome, said “when we speak to labs, they always highlight data management, analysis, and interpretation as key challenges facing the field." To meet this challenge, "our software development is focused on increasing automation, improving workflow, and helping labs to easily interact with public databases and other data sources," he said.
Part of the reason that users are increasingly focused on the interpretation of array results, rather than the attributes of the different platforms, is because some consensus exists in the cytogenetics community about what the optimal content and format for chromosomal arrays should be.
David Ledbetter, director of the division of medical genetics at Emory University School of Medicine in Atlanta and head of ISCA, noted that while in the past vendors jockeyed for market share on the basis of whether their platforms contained primarily copy number variants or SNPs, all four major manufacturers — Affymetrix, Agilent Technologies, Illumina, and Roche NimbleGen — are moving to include CNV and SNP content in future iterations of their arrays.
Both Affy and Illumina entered the market with their SNP genotyping platforms, which also contain CNVs, while Agilent and Roche have offered comparative genomic hybridization arrays, which enable users to survey CNVs. According to Ledbetter, all firms are now developing new products that will enable users to access both forms of variation.
"There is a slight difference between the arrays with SNPs and without SNPs," said Ledbetter. "At the moment, the copy-number-only platforms have a bit cleaner performance, higher signal-to-noise ratio, than the SNP arrays," he said. "Since the primary interest of the test is copy number, we [at Emory] use the copy number arrays, but we would like to get the SNP information at the same time."
Ledbetter said that the "SNP platforms are improving their copy number detection capabilities," while manufacturers of copy number platforms are now adding SNPs, "so those will converge at some point."
He said that "everybody wants" to be able to look at copy number and SNP content, as only SNPs can pick up conditions such as uniparental disomy, a condition that occurs when a person receives two copies of a chromosome, or part of a chromosome, from one parent and no copies from the other; and consanguinity, when the parents of a patient share common ancestors.
According to Ledbetter, the convergence of the array platforms is "moving pretty quickly" and "by early next year, there will probably be several improved versions where the copy number and SNP detection capabilities meet our wish list."
Having those needs met has offered users more time to focus on solving the issue of interpretation. Ledbetter stressed that it is "not a major issue," but one that ISCA aims to address.
"I would say that 95 percent of all results are clearly normal or clearly abnormal," said Ledbetter. "Out of the remaining 5 percent, by doing parental studies, we are left with only several percent that we are still uncertain after we have done all the tests that we can, and we may or may not have a guess of whether it is pathogenic or benign."
While "no laboratory-developed test or diagnostic is perfect," Ledbetter noted that clinicians are sometimes troubled by the inability of array-based genetic testing to explain everything. "It is true that every lab director and clinician feels concern and sympathy for patients when they get an ambiguous result," said Ledbetter. "They would rather that number be zero."
To help them better interpret results, the US National Institutes of Health last year awarded Emory a $3.5 million grant to develop a CNV atlas that would aid ISCA members and other users in making a call on whether an array finding was benign or pathogenic (BAN 10/6/2009).
One issue that ISCA members discussed during last week's meeting was that data is not being uploaded frequently to the ISCA database or other public databases, as it takes time to de-identify the sources of the data and to curate it to meet the specifications of the database.
Ledbetter said that ISCA now provides protocols and kits for labs that wish to contribute retrospective data from existing studies, as well as more complete data in a prospective manner after receiving approval from the requisite institutional review boards. He also said that vendors could aid in helping customers upload their data.
Marilyn Li, a geneticist at the Tulane University Medical School's Hayward Genetics Center and ISCA member, said the database would give users the necessary genetic information and the phenotypic information in order to make a call on a case. "That will be very helpful for clinical diagnostic laboratories," Li said. "The ISCA database will have results categorized as pathogenic, benign, and variation of clinical significance," she said. "We hope in the future to sort those cases into one of the first two categories."
Ledbetter predicted that in the next two to three years, clinicians' anxiety over uncertain results is going to be "greatly reduced" as a result of ISCA's data-sharing plan and database. Still, he said that companies still have a role to play in helping users reach a diagnosis.
"Customers don't think software development is going as fast as it could," said Ledbetter. "It seems easy to us to add some user-friendly features to make our lives easier, but it takes time for the companies to implement software changes."
Related to software is the issue of providing ways to make internal lab databases "communicate with central databases like ISCA, upload the data, and receive that data back with search tools to inform your interpretation," Ledbetter added. He said that ISCA is now putting customers and vendors together to work on such software functionality.
James Clough, vice president of clinical and genomic solutions at Oxford Gene Technology, said the company is working on new software tools to help its customers. The British firm last month announced that its CytoSure Interpret software can transfer array CGH aberration data directly into Cartagenia's Bench constitutional cytogenetics database and analysis platform.
OGT said at the time that the new capability will provide cytogeneticists with additional data management options such as data sharing within, and across, various research facilities, and support for genotype-phenotype correlation analysis.
"When we first started, our software was really for visualization of results," Clough said. "Now it's an analysis tool, and we have partnered with Cartagenia, which offers a phenotype software tool." The "key question" facing OGT's customers is "getting clarity on what is causative, what is benign, and what is of unknown significance." Time is also a factor, Clough said.
"The clinician wants to know if it is causative or not," said Clough. "The big challenge is how you, in a manageable way, spend your time on interpretation," he said. "You could spend ages on every case. Now that we have got high-quality array platforms and reproducible results, I think [the main challenge] is the interpretation."
Toward Better Methods
KU Leuven's Vermeesch said he has been using OGT's CytoInterpret software and Cartagenia's Bench platform in his lab, but cautioned that just having access to data is not going to be enough to close the interpretation gap in array-based cytogenetics.
"The problem with the databases is that you check your findings and see that, 'Well, my colleagues have called it a pathogenic mutation, so I will call it a pathogenic mutation,' but that is not evidence based," said Vermeesch. "You need more resources to mine that data to give a better interpretation," he said.
Emory's Ledbetter said that the community needs to design an evidence-based process and establish the criteria for declaring some variants pathogenic and others benign. "Right now, we have a small group of experts that are trying to pilot this and formulate a strategy, but we are going to need a lot of community input, and it is going to be some kind of refereed, wiki process," said Ledbetter.
The group of experts developing the protocol for evidence-based reviews of data include clinicians and bioinformaticists, but Ledbetter expressed a need to open the discussion to a "broader community of clinicians and lab experts, frontline clinicians, and clinical labs and get their input and see what they want us to do, where we need more uniformity, and where we need new developments."
Michael Watson, executive director of the American College of Medical Genetics, said that ACMG is developing a web-based educational program on how to interpret array results. He noted that findings of unknown significance are a "major issue" for the US Food and Drug Administration, which is looking to regulate the way arrays are used in cytogenetics. Watson said that ACMG's cytogenetic array interpretation program is in development, but once finished could be made available at all of ACMG's training programs for free.
"When people ascribe to certain ways of interpretation, they are more likely to come to similar results," Watson noted.
For those charged with actually consulting patients, inconclusive findings, while still only about 5 percent of cases, pose a special challenge. While clinical geneticists that have adopted the technology argue that genetic testing has always yielded some ambiguous results, some say the promise of new technologies to answer all questions has fostered some skepticism.
"I don't think it's that much different from the genetic testing that clinicians have always used, but because it's a new technology and it is bringing to the forefront this idea that you are finding things you cannot yet explain, it has become a rallying point for clinicians that are afraid of the unknown," said Darrel Waggoner, an associate professor of human genetics at the University of Chicago.
"They say they are nervous about the new technology in spite of its potential, and those kinds of attitudes are an obstacle to the adoption of the technology, but I think that with education, familiarity, realization of the benefits, those attitudes will all be overcome," he said.
David Miller, a clinical geneticist at Children's Hospital Boston, said that better interpretation tools and methods will smooth the adoption of arrays in cytogenetics.
"Geneticists are becoming comfortable with the idea of ordering a test where the results are unclear now, but maybe in a year or two there will be more information available in the literature that will shed some light on the findings," said Miller. "We are comfortable telling people that we can't explain this finding now, but we may be able to later."