Skip to main content
Premium Trial:

Request an Annual Quote

Genetic Testing Firms Tout Underlying Bioinformatics Capabilities as Key Business Component


By Uduak Grace Thomas

Last week, Gene Security Network, a prenatal genetic testing firm based in Redwood, Calif. launched a non-invasive prenatal paternity test.

Separately, Everist Genomics, a biotechnology company based in Ann, Arbor MI, launched a real-time PCR-based test dubbed OncoDefender-CRC that identifies early-stage colorectal cancer patients that may be at risk of the cancer recurring following treatment.

While their tests may target different segments of the genetic testing space, officials from both companies told BioInform that their underlying bioinformatics capabilities — both of which were developed in house — form a key component of their respective businesses.

Peter Lenehan, Everist’s chief medical officer, told BioInform that his firm, which was formed in 2002 and currently has eight employees, aims to "develop prognostics with a bioinformatic bent."

"Our company is based on bioinformatics where we use molecular biology data, protein data, gene expression data, SNP analysis, and other DNA data, bring that all together … analyze that data, make a story out of it, and get some results," he said.

To that end it has launched OncoDefender, a colorectal cancer diagnostic test that was created using machine learning techniques, and it is currently working on a second test that relies on a separate internally developed algorithm to assess a patient's risk of developing cardiovascular disease.

OncoDefender analyzes patient tumor samples for the expression of five specific tumor genes and then uses a genetic programming algorithm to calculate a relative recurrence score for each patient that stratifies him or her into low- or high-risk categories for assessing the likelihood of the cancer recurring.

The test provides a way to identify which patients in the early stages of colorectal cancer should receive therapies on top of surgery to clear any cancerous cells missed by the scalpel, Lenehan explained, adding that there is an unmet need for such tests.

He said that in the US, about 150,000 patients are diagnosed with colorectal cancer yearly. Of that amount, about 60,000 are in the first two stages of the disease. These patients are usually treated with surgery that may be followed with chemotherapy to clear away residual cancerous cells.

However, following surgery of stage one cancers, "the assumption is you have got all the tumor out" and as such "it's not worth giving chemotherapy," he said. The reality, however, is that "one in seven patients actually recur … and over half of those patients will succumb to their disease.

"To us that’s a significant number of patients when you combine that with the number that recur in stage two … that could benefit from adjuvant chemotherapy which, if effective, could get rid of those small tumors cell which are left over after surgery," Lenehan said.

Everist estimates that the cost of missing high-risk patients who subsequently develop advanced cancers comes to $250,000 per individual; while the cost of running OncoDefender and providing adjuvant therapy following surgery would cost less than $50,000.

Compared to similar products, OncoDefender's approach to testing is more "definitive," providing physicians with more useful information because it has a higher positive predictive value — or the proportion of patients with positive test results who are correctly diagnosed — than the competition, Lenehan said.

OncoDefender is "dichotomous because it gives a yes/no answer or a high-risk/low-risk answer," he explained, noting that some competing tests are "continuous," reporting certain percentage risk without indicating the significance of that percentage.

Furthermore, while other tests such as Genomic Health's Oncotype DX test and Agendia's ColoPrint test, which focus on stage two colon cancer, OncoDefender encompasses both the first and second stages, "so we have a larger population to draw from," Lenehan said.

Everist promises to complete its analysis and send a report back to physicians 7-10 days after it receives the tumor sample.

"What they do with this is up to them," Lenehan cautioned. "We are not claiming that the tumor will respond to chemotherapy but we are telling them that … if this patient had a high-risk tumor, then adjuvant chemotherapy should be highly considered as appropriate … and this patient could potentially benefit from it. … If its low-risk then it would imply that perhaps watchful waiting would be a better management route."

OncoDefender is priced at $3,300 per test, which Lenehan said is comparable to similar tests. The cost is usually covered by insurance; however, Everist offers a financial assistance program that can cover up to 80 percent of the price for individuals without insurance.

Replacing Invasive Prenatal Tests

GSN's proprietary system, called Parental Support, uses a combination of bioinformatics algorithms and information on single nucleotide polymorphisms to detect small quantities of fetal genetic data in a sample of the mother's blood.

Matthew Rabinowitz, GSN's CEO, explained that its newly launched test replaces invasive and often unreliable screening tests that could be hazardous to pregnancies.

For example, he told BioInform that some non-invasive tests miss about one in six cases of Down syndrome, while alternative invasive tests could cause miscarriages.

"We developed this technology platform [to] give expecting parents a test that is really reliable and really safe," he said. "We do this by combining innovations in biomedical research with some advanced bioinformatics and statistics."

The test requires a sample of a pregnant woman’s blood, Rabinowitz explained. Depending on the amount of fetal DNA present in the woman's blood, a sample from the father, either in the form of a blood draw or buccal sample, may also be required.

As a next step "we take data from the Human Genome Project such as the HapMap database," which contains data on human genetic variations and "we use that a priori information from the parents and from the human genome project to [build] a model that gives us much more statistical power with which we can look at … the fetal DNA signal" and separate it out from the maternal signal, he said.

"One way of looking at it is you are doing a Bayesian maximum likelihood analysis," he explained "You have data and you want to find the maximum likelihood hypothesis given the data."

For instance, two hypotheses that could explain a chromosomal abnormality such as an aneuploidy may be that there are two chromosomes from the mother and one from the father, or two chromosomes from the father and one from the mother, instead of one per parent.

In this case, “you are trying to find the … maximum likelihood hypothesis given the data that you've measured," Rabinowitz said. With Parental Support, "not only do we say this is the most likely hypothesis, we will also say this is the amount of confidence that we have that this is the right hypothesis and typically our confidences are well above 99 percent," he said.

Parental Support will have to compete with products from companies like Sequenom, which plans to offer a prenatal test that relies on shotgun sequencing.

These tests "count the number of reads that you see on chromosomes" and are set up as "single-hypothesis-rejection tests," Rabinowitz said.

They "look at the number of sequence reads that you see on a chromosome … and compare that to the number of reads that you expect to see if that chromosome was normal." he explained. "If that chromosome has more reads than you would expect to see, you would call that chromosome abnormal."

In other words, "you set some threshold which gives you some false positive rate … and anything that is above that threshold, you will call [abnormal]," he said.

The problem with this approach, according to Rabinowitz, is that it is unreliable for very small quantities of fetal DNA and could miss potential abnormalities that don't exceed the set threshold.

"It is a very flaky approach when you have less than 10 percent fetal DNA [for example]," he said. "If you want to do testing in the first trimester as we currently do, about 30 percent of the cases will have less than 10 percent fetal DNA."

In addition to its paternity test, GSN's algorithms have been used by in vitro fertilization centers to diagnose genetic mutations in the developing cells as well as chromosomal structural problems that could affect the health of the pregnancy, Rabinowitz said adding that the firm's approach is comparable to "metaphase karyotyping" which he described as the gold standard for chromosome testing when samples contain many cells.

Parental Support has also been used to analyze "products of conception" following a miscarriage to determine its cause. Additionally, it clarifies "30 percent of ambiguous results" that are associated with regular karyotyping approaches, he said.

There are slight differences between the way the algorithm is implemented in the tests, Rabinowitz noted, "but the underlying statistical concept is pretty much uniform throughout."

GSN plans to launch a non-invasive prenatal diagnostic test based on its informatics platform for Down syndrome and other genetic disorders, and plans to assess it in a clinical trial backed by the National Institutes of Health and with input from research centers around the country.

GSN has licensed the test to the DNA Diagnostics Center, which is distributing it under the moniker DDC's Non-Invasive Prenatal Paternity Test with a $1,625 price tag.

Moving forward, Rabinowitz said that the company is developing more non-invasive prenatal tests similar to Parental Support, "and bioinformatics will have a role in that."

"We will be the best technology for doing a reliable, safe diagnostic that helps the couple know that they have a healthy child or gives them the most reliable information as fast as possible," he said.

Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com