With the opening of a new manufacturing facility in Tempe, Az., this fall, and the introduction of new rat and human chips, Motorola Life Sciences has been quietly ratcheting up its microarray manufacturing capabilities.
Although the facility is strictly off-limits to members of the press, BioArray News has gotten a picture of the companys manufacturing process through interviews with Motorola employees.
Overall, this picture is one of a company that has sought to enter the high-density array field as a careful conscientious follower. Motorola wants to make a high-quality alternative to the Affymetrix offerings, rather than be a bold innovator willing to risk everything on an entirely new platform. Its 10,000-spot human arrays aim for consistency and reliability, while its rat ADME array and p450 array seek to fill well-worn niches. By the end of the year, the company is planning to release a 2,600-spot human SNP array as well.
In the matter of actually making the arrays, the first decision was content selection. Like Agilent and now-departed rival Corning, Motorola chose to go with Incytes genes and ESTs. But it also combines this content with public domain sequence, covering Affymetrixs fishing grounds.
Like Affymetrix, the company also chose oligonucleotides, not cDNAs a smart move given that this is where most experts believe arrays are going and also one that hews close to the path broken by Affymetrix.
To design the oligonucleotides, Motorola chose to use its own algorithms, and then to validate these oligonucleotides against a variety of tissues. For its rat ADME chip, the company used the services of Compugen, which has developed algorithms for oligonucleotide selection that not only minimize cross-hybridization, but account for splice variants.
These oligonucleotides are manufactured by an undisclosed outside supplier, then are shipped to the manufacturing facility, where they undergo quality control testing, said Nancy Schmelkin, the marketing manager for Motorolas biochip products.
Next, the oligos move to the companys large biochip manufacturing area.
This facility gives us enough space to do thousands and thousands of chips, said Schmelkin. There is not a customer that we could not take on.
CLASS 10 CLEANROOM
The arrays are deposited using Packards piezoelectric non-contact deposition robots. The slides for the arrays, which are coated with the companys proprietary porous hydrogel, see a class 10 environment, Schmelkin said. This is the highest of standards for clean rooms, and means that there are less than 10 particles of dust particles 0.5 micrometers or larger per cubic foot of air.
During the manufacturing process, the company uses serial numbers and barcodes. Every step is serial-number tracked, so anyone doing quality control can match each lot of oligonucleotides to operations on the machine. During the process, quality control is performed on every chip and every spot. This might involve not only making sure that the oligo was there, but whether there is a contaminant such as a speck of dust, said Schmelkin.
In a final step, the arrays are packaged in sets of six, which are then grouped in a larger plastic case that contains 24 microarrays, the minimum quantity sold to customers.
Customers are shipped the arrays along with a hybridization protocol and reagent kits, which have received as much care and thought as the actual manufacture of the chip itself. There is one thing, however, that is missing from Motorolas kit: Information about the actual sequence of the oligonucleotides spotted down on the chip. The actual sequence we consider confidential. And most dont want to know the sequence of all ten thousand genes, said Schmelkin.
In this decision, Motorola is perhaps sticking to the Affymetrix microarray marketing manual a little too closely. Affymetrix has been repeatedly criticized for its failure to provide the sequences of its oligonucleotides, and decided this fall, as part of its new Perestroika, to reveal these sequences in new software planned for early next year.
But Motorola believes that customers really care about the specificity and sensitivity of the array more than the sequence. And on this point, it is all too willing to provide information. These arrays have coefficients of variation that range from 12 percent to the upper teens, and on the smaller arrays such as the rat ADME chip, where there are three nucleotide probes dispensed for every target sequence, the coefficient of variation drops to as low as six to seven percent.
To demonstrate the prowess of its platform, Motorola Life Sciences has been running a proof of performance program, in which customers can send in their sample, and the company will hybridize the sample to the arrays, then send back the data and allow customers to compare this data to other platforms.
This tactic, like the companys overall conservative strategy to stick with the spotted array on a slide, rests on a major assumption: the belief that users of Affymetrix and other platforms are dissatisfied enough to try something new, but dont want something too different. And, while the company may not have wanted to go out on a limb for a new and untested technology, this assumption that scientists want more of (almost) the same thing could turn out to be a major gamble.