A new agreement between Australia's Peter MacCallum Cancer Center and National Information and Communications Technology Australia highlights the center's growing need for statistical expertise to handle its array data and demonstrates the consequences that higher density microarrays can have on researchers, according to officials involved in the project.
Adam Kowalczyk, who heads a bioinformatics research group at NICTA's office in Canberra, told BioArray News this week that his organization and Peter Mac signed the agreement last month, and that NICTA will upgrade the informatics capabilities of the center. He said that Peter Mac's capabilities are being strained by a growing number of projects coupled with more data generated by higher-density research-oriented arrays.
"The whole project can actually be divided into 10 different projects; this isn't one project, this is a bundle," Kowalczyk said. He added that the cancer center, which employs more than 200 scientists, is strong on biology but weak in the IT department. "They haven't got anyone that takes care of IT," Kowalczyk said. "There is a strong need for statistical expertise."
It's a need that's being exacerbated by the influx of high-density microarrays, mainly from Affymetrix, according to officials from Peter Mac, that are replacing the older arrays due to the large amount of content available. The center's research director, David Bowtell, told Australian Biotechnology News last month that Peter Mac has been accumulating datasets involving hundreds of samples with detailed clinical annotation, from relatively basic analytical techniques.
The increasing density of Affymetrix GeneChips would pose a "massive problem" in pattern recognition at the center.
Bowtell said that the cancer center already had 250 annotated samples from early, 10,000-element arrays, and 230 samples from ovarian cancers analyzed with 58,000-element arrays. He said that the increasing density of Affymetrix GeneChips would pose a "massive problem" in pattern recognition at the center, according to the news report.
Bowtell cited Affy's GeneChip Exon 1.0 ST array, launched in October 2005 with more than 5.5 million elements, as an example of the higher-density chips his center is using (see BAN 10/5/2005).
Affymetrix also launched its GeneChip Mapping 500K Array Set that month, which it claims is capable of genotyping on average 250,000 SNPs for whole-genome association studies. Also, last month Affy promised to roll out 11 distinct high-density tiling array products over the course of 2006 (see BAN 12/14/2005).
Alicia Burt, Affy's senior product manager, told BioArray News at the time that one of the products, the Human Tiling 2.0R Array Set, will comprise seven tiling arrays with approximately 45.5 million total probes on the set.
Bowtell did not respond to e-mails or phone calls from BioArray News seeking comment by press time, but Kowalczyk agreed with this assessment of the informatics dilemma created by the advent of higher-density chips.
"It's very recognizable as a problem that needs to be addressed. It definitely requires attention," Kowalczyk said.
Averting a 'Mess'
Would buying software off the shelf, instead of commissioning customized software, help solve this problem? Kowalczyk, a former cancer center employee, said that Peter Mac had distinct informatics needs that weren't being met by commercially available software. He said that the center had been using commercially available software, but multiple analysis obstacles coupled with multiple platforms were slowing down its data-analysis capabilities.
"Peter Mac is using mostly [Agilent's] GeneSpring and some other packages ... but it is not sufficient," Kowalczyk said. "They are good for preliminary analysis, but when you go to do something more systematically, it's not good enough. Once you start dealing with a situation where there's a lot of data and multiple software packages, it can turn into a mess," he said.
"It's better to write customizable software that addresses specific problems," Kowalczyk said. The new software developed by Kowalczyk's team at NICTA will allow Peter Mac researchers to use predictive statistical models on microarray data to diagnose cancers. Kowalczyk said that the project is ongoing and objectives may change over time.
Large cancer centers like Peter MacCallum aren't the only organizations that have opted for customized data management tools. For example, Petra Ross-Macdonald, a senior research investigator at Bristol-Myers Squibb, told BioArray News in August that BMS had opted to build its own system rather than buy commercially available software to handle its microarray data (see BAN 8/31/2005).
"Once you start dealing with a situation where there's a lot of data and multiple software packages, it can turn into a mess."
"Every big lab has a lot of microarray data and they all choose to handle it differently," Ross-Macdonald said at the time.
However, two US-based companies, Partek and Stratagene, have software packages available to handle the loads of data that are being generated by high-density chips like Affy's Exon Array. And, moreover, they are seeing increased demand despite the tendency of some labs to opt for customized software.
According to Tom Downey, CEO of Partek, which sells its Genomics Solution software for array data analysis and management, "sales [have been] increasing at academic labs and at the large pharmaceuticals as well as a result of the newer high-density arrays."
"The large pharmaceuticals are a little bit slower to change, because they have a larger investment in their existing systems and methods," Downey told BioArray News last week.
However, Downey said the need for better software tools that centers like Peter Mac are experiencing may grow as more labs adopt higher density chips.
"I think the migration to the new higher-density arrays is inevitable, as they simply contain more meaningful information than the previous generation of microarrays," he said last week.
Meanwhile, Stratagene has also developed a platform to handle data from Affymetrix's exon arrays. The company released a statement in December that it will make an ArrayAssist Exon module available for its software this month.
Stratagene did not respond to an e-mail seeking comment on the ArrayAssist Exon module by press time.
Still, David Weber, senior vice president of marketing at Stratagene, told BioArray News' sister publication BioInform last month that additional ArrayAssist modules for tiling arrays, 500K SNP arrays, and other upcoming Affymetrix arrays will roll out in "the early part" of 2006. Last week Partek's Downey said that Genomics Solution can handle the new Affy tiling arrays and pledged that Partek will "continue to deliver meaningful statistical analysis and visualization tools for these ... new technologies in addition to integrating the information from the various platforms."
-- Justin Petrone ([email protected])