Using squeaky-clean ethics and strict standards, Genomics Collaborative and Ardais lead in using IT to bring human tissue banking into the era of clinical genomics
March 17th, 2004
Mark D. Uehling
A BIOMARKER for a diagnostic test for colon cancer. That was what the customer needed to validate, with help from Genomics Collaborative, a Cambridge, Mass., company selling tissue, blood, and associated genomic data.
Michael Pellini, the company's CEO, explains what happened next. "What the company had to do," he says, "was not study another 50 patients. They needed hundreds of the exact same type of colon cancer collected in the same fashion. In two separate projects, we were able to deliver that."
Genomics Collaborative needed a month for 250 specimens in one instance, and two months to find 350 samples in another. Rounding up the specimens using traditional methods — waiting for them to walk through the door of one isolated doctor's office or hospital — could have taken 18 months or longer.
Needing tissue is not unusual. What's new, in an era of heightened privacy rules, is demand for hundreds of related tissue, blood, and DNA specimens. Such samples must be tracked and managed with industrial-strength software. They must be linked to reams of anonymized clinical data. Companies like Genomics Collaborative, Ardais, and Gene Logic are trying to make a business of filling such orders, which academic centers struggle to fill. Why? Even the largest university encounters only a trickle of patients with a particular disease or genetic trait.
"There is a growing use of human tissue samples all the way from early discovery into preclinical and clinical work," says Martin Ferguson, co-founder and senior vice president of bioinformatics at Ardais, another leading clinical genomics company. "Everyone wants to do personalized medicine; everyone wants to do molecular medicine. And ultimately, the best model for human disease is ... humans."
The tendency in academia, all too often, has been to collect little clinical data about such specimens, or to use software and records slapped together as a summer project. Among leading repositories, only the Mayo Clinic now uses bar codes to track samples (see Tissue Repositories).
Surprisingly, academic and government tissue banks are only just starting to consider all the intricacies of the challenges of tissue banking in a genomic era (see "Study: Federal Banks are Lagging"). Some of the challenges are political or logistic: Is every lung biopsy going to be handled exactly the same way, processed and frozen within the same number of minutes? If so, some surgeons and pathologists will have to change their beloved procedures. If not, the data may not be strictly comparable. Other challenges are informatic: Has the terminology in the tissue database been standardized?
All too quickly, most of the issues in tissue banking in the genomic
era transcend parochial institutional fiefdoms. This is not an area where
pathologists in academia are showing the way. There's no better proof of
that than the bankruptcy of an early entrant to the field, Impath. The
company was founded in 1988, assembled a large network of pathologists,
and amassed tissue and data from more than a million patients. Impath declined
to articulate its present strategy.
Founded in 1994, Gene Logic is one of the oldest existing companies offering a combination Putting Tissue on a Slide
Gene Logic's revenue is rising, but the company has burned through at least $159 million.
of tissue and data (see "Putting Tissue on a Slide"). But since the late 1990s, Genomics Collaborative and Ardais have been delivering specimens for drug discovery. Both companies already use national or international networks of physicians and hospitals and — more importantly — clear, precise, uniform informatic standards for collecting, preparing, and tracking specimens.
For Genomics Collaborative's Pellini, a physician and MBA, the tissue isn't even the half of it. Which is not to say quantity doesn't matter: The company has collected 500,000 genetic, blood, and tissue specimens from 120,000 patients. Tissue donors' identifying personal data is stripped away, easing privacy concerns. The company is drawing from a global network of 600 physicians in order to ensure that if it needs a particular stage of prostate tumor, no single site will feel undue pressure to bend Pellini's rules.
He, in turn, knows a network of collecting institutions ensures he can deliver specimens and data faster than any contract research organization or the largest academic medical centers. That's when he can help pharma, he believes. "It's when they have found something interesting, and they want to go to 20 or 30 or 300 or 2,000 or 3,000 specimens. That's where we play."
Genomics Collaborative is unique in allowing customers to triangulate between tissue, blood, and genetic data. The company doesn't advertise itself as doing systems biology per se. "We say, the integration of human genetics, functional genomics, and proteomics. That's where you understand the pathway. That's where you understand if you perturb something at the genetic level what happens at the protein level. Unless you have these different components, you don't know."
SCRUBBED DATA: Michael Pellini's Genomics Collaborative sells
blood, tissue, and genomic data while maintaining strict standards.
Genomics Collaborative collects all its specimens according to protocols that have been approved by institutional review boards (IRBs) and according to the practices customary in clinical trials. Those sorts of tiny details are hardly afterthoughts; they are what the company incorporated into its workflow to maximize the subsequent use of tissue and searching of its attendant data.
Unlike Ardais, which also caters to academic groups, Genomics Collaborative has geared its operations to the most demanding customers. "Everything is double entered," Pellini says, referring to the redundant entry of information in clinical trials. "Everything ends up in an Oracle Clinical database. We scrub the data. When there are holes in the clinical data, we are able to go back to the sites to retrieve the missing data. If we can't, or there are inconsistencies, we terminate the specimen."
Pellini points out that the custom-built informatic system at the company also logs a variety of other images and documents that are part of every patient's case. "If we have a sample from a patient with coronary artery disease, yes, on the case report form they will write in the percent stenosis of the artery. If we are looking at a specimen from a 45-year-old female with coronary artery disease, we can pull up all her clinical information and we can see the actual radiology report, the actual angiogram. That's viewable."
That sort of obsessive attention to detail allowed the company to sign deals with 18 companies in the second half of 2003, including Novartis, Roche, Pfizer, GlaxoSmith-Kline, Johnson & Johnson, and others. Despite such prominent customers, the GQ-worthy Pellini says the business is anything but glamorous: "There is nothing too sexy about this. It's dotting the 'i's, crossing the 't's. It's just making sure you have all the pieces in place."
The privately held company has deep ties to Harvard, and originally wanted to pursue drug discovery on its own. But eventually Genomics Collaborative recognized — perhaps like J. Craig Venter on a visit to Las Vegas — that certain casino tables are only for the very deepest pockets. However vast its resources, Genomics Collaborative lacked the cash to find new drugs and push them through the gauntlet of clinical trials.
So its investment in drug discovery infrastructure was re-engineered into a platform for clinical genomics. The company developed its own laboratory information management system (LIMS), and hooked up everything to Oracle Clinical. Pellini likes to talk about the "tens of millions of dollars" his company sank into developing its IT system and the automated sample handling stations, the freezers connected to pagers in case of a power outage — not an uncommon thing in funky Cambridge, Mass., as it turns out. Naturally, like Ardais, the company has extensive backup technologies and procedures in place.
As often as not, Pellini says, he is getting business by word of mouth.
"What we're proud of, and what's important to us, is we don't have to work
with companies in a cookie-cutter fashion. If it's the functional genomics
group that needs access to certain tissue types, you know what? We can
do that. If it's the functional genomics group that needs access to DNA
specimens because they want to cross over into human genetics, we can do
that. If it's a group involved in a clinical trial, and they want to look
at protein biomarkers and serum samples, we can do that. We're nimble enough
to work in these different areas."
The story is just as interesting a few miles away, in Lexington, Mass., where freezers at Ardais hold 220,000 samples from more than 15,000 donors. Ardais has also invested in its own standard operating procedures, its own software, and its own IRB-approved protocols.
Ardais, too, has been doing well, perhaps because most U.S. surgeons and pathologists neither know what it costs to collect a sample, nor have the IT systems in place to offer leftover tissues to corporations.
Ardais says it signed up 15 new customers in the fourth quarter of last year, for a total of 40 during the year, including Biogen Idec, Celera Genomics, Corixa, diaDexus, and Epigenomics. That's a nice run in light of having hired two CEOs in the past 18 months.
Ardais is drawing samples from 300 physicians and four academic sites; it has suspended tissue-collection efforts at Beth Israel Deaconess Medical Center in Boston due to what Ardais says is a lack of demand for banked tissue by that hospitals' researchers. But even in the latter case, clinical data about the samples is still being gathered, and Ferguson says the samples remain precious. "We believe these will become more valuable with time, as we start to learn the outcomes of the patients. "
The company is beta-testing its software at one academic center that is not supplying it with tissue, and says that licensing its tools to manage specimens could become its own business. Annotating the specimens as they are logged into the database greatly facilitates the searching later, as pathologists are notoriously prone to having six different names for the same cancer.
Says Ferguson: "We annotate it in a set of standardized vocabularies, so if one physician says 'prostatic adenocarcinoma' and another says 'adenocarcinoma of the prostate,' they actually get the same code in our database." Just using a Google-type search on the data could be very frustrating without such preparatory work, Ferguson says: "You would have to do a lot of different searches to come up with the set of examples that might or might not be of the type you want."
As he outlines the little-known vagaries of human specimen collection,
he points out that a scientist, surgeon, or pathologist in government or
academia traditionally drops a small portion of a tumor into a Ziploc bag,
scribbles a few words on the bag with a marker — and tosses it in a freezer.
Slicing the Salami Thin
At Ardais, this routine has been industrialized with bar codes, FedEx, and standardized containers. Once adequate tissue samples for each patient's future care have been set aside, there is still plenty of tissue left to ship to Ardais. The participating collection universities immediately subdivide each specimen into as many as 10 to 15 slices or chunks. Ardais can also reserve adjacent normal tissue at the same time, an invaluable aid for some projects. The bottom line is that Ardais gets more out of the same sample. "On average," Ferguson says, "Ardais gets 10 to 15 specimens per case as opposed to one."
That allowed Ardais to recently send a top 10 pharmaceutical company more than 1,500 samples on a single order. You can talk about tissue all you want, he says — and Ferguson is happy to. But statistics are what really underpin customer demand for tissue. "People have come to realize that the quality of the research project is highly dependent on how well the samples were collected and annotated in the first place," he says.
"Before, in 1997, 1998, we knew about 4,000 genes," Ferguson says. "Now
we can simultaneously analyze 40,000 genes in a heartbeat. When you start
using a large number of data points like that, you need a large number
of samples. The only way you're going to get that is to have a lot of collection
Copyright © 2004, Bio-IT World Inc.