It has been over a month since Affymetrix marched out its latest U133 array set, with the whole draft assembly of the human genome squeezed onto just two chips.
But while the confetti and ticker tape has long been cleared from the parade route, few array users have used this new product enough to deliver a definitive verdict, according to an informal (and decidedly unscientific) survey of users conducted by BioArray News. Many have decided to finish all experiments with the previous-generation U95 human chips before getting started on the new arrays to avoid the technical headache of comparing results from the new chips to the old chips within the same experiment.
“The Affymetrix U133 arrays seem to be doing what they are supposed to. I think it’s great that Affymetrix is giving you more genetic content for your dollar. But for us the jury is still out,” said Andrew Brooks, director of the functional genomics center at the University of Rochester Medical Center.
No Noise is Good Noise
This lack of noise about the new chips could be seen as good news for the company, considering the past rumblings of activist Affymetrix users. The company’s disclosure of probe sequence on its NetAffx website — in response to longstanding criticism about the secrecy of these sequences — and its new algorithm seem to have muted criticism for the time being.
“I think [the new chips and sequences on NetAffx] are continuing to demonstrate Affymetrix’s responsiveness to market forces,” said Mark Geraci, who directs the microarray core at the University of Colorado at Boulder. “Just when you get frustrated with their product they do something that appeases the masses, and actually this one is quite thoughtful and quite helpful.”
Those who have thoroughly test-driven the chips say they are pleased with the results. Gene Logic, which is the largest user of Affymetrix chips and runs through about 2,000 arrays per month, has delivered a glowing initial review. “We aren’t finding that we have any problems with the U133 arrays. Technically they are running fine,” Doug Dolginow, the company’s senior vice president for pharmacogenomics, said last week. “It’s clear that Affymetrix has really now moved into the commercial scale where these chips come out and tend to work because they are quality-controlled quite well, whereas four or five years ago there was not the same scale and reliability.”
Minor Grumbling Over NetAffx, Algorithm
Affymetrix user grumbling, however, has not stopped entirely. Some users complain that NetAffx is not user-friendly and has some rough edges. “I personally don’t think that the design of the website is that fantastic — it’s difficult to navigate — and for that reason I don’t use it that much,” said Brooks. Another criticism is that some U133 probes do not have annotations indicating what gene they represent. But these problems can be easily fixed and don’t negate the site’s usefulness, users agree.
“The ability to get at the oligo sequences for us has been a very big help,” said Geraci. “But the site itself doesn’t run entirely smoothly. It’s just sort of a webmaster’s problem.”
Another grumble concerns the company’s new data-mining algorithm. While many users have praised the company for delivering a statistically based improvement over the previous algorithm, “There are still problems with the new algorithm, and a number of people are contemplating writing their data extraction algorithms for Affymetrix chips,” said Peter Tolias, director of the Center for Applied Genomics at the International Center for Public Health in Newark, NJ. “The new software removes the negative values when the mismatch hybridization is higher than the perfect match. They make it positive, and in doing so they mask the raw data. People want to know what the raw data is and make their own judgment as to whether a call is real or not,” Tolias said.
How to Compare U95, U133?
Complaints aside, users are chewing their pencil nubs over the problems involved in comparing data from the new chips and new algorithm to data generated with the old chips and old algorithm.
“People have to take a step back,” said Brooks. “We have to reevaluate how we assess the quality of our arrays based on the output being different as a function of a new array type and design, and the new algorithm.” While NetAffx does ofter a visual comparison of the U95 and U133 probes, Brooks insisted that “you cannot actually truly compare the data between the two arrays yet” on NetAffx.
Mark Watson, director of Washington University-St. Louis’ Affymetrix core facility, plans to resolve the comparison issue with “help from Affy and a bit of bioinformatics muscle. We won’t compare directly between the two chips, but we will compare results of experiments run on U95s with new experiments run on U133s.”
Others plan to do a test experiment on both chips, and then compare the data to see what kinds of changes there are on comparable probes. “I assume that Affymetrix must have done this experiment, but a lot of people are asking [about these results] and no one knows the answer,” said Greg Khitrov, director of Rockefeller University’s microarray facility.
Even with this comparison challenge, there is general agreement that Affymetrix’ new offerings represent progress. And if imitation is really the sincerest form of flattery, Affymetrix has recently received a major compliment from Qiagen Operon. The company recently launched its own NetAffx-like web-based search tool, the Operon Microarray Database.