Microarrays measure up well

To address concerns about microarray reliability, several groups have attempted to evaluate and compare different platforms, but these efforts themselves are open to question. "Each one of those studies was biased one way or another," says Winston Kuo, Harvard researcher and lead author of a recent cross-platform study.¹ Kuo points to flaws in earlier efforts, such as different protocols, nonstandard preprocessing, and reliance on probe annotation information.

Since each vendor uses a different set of probes and method of analysis, comparing platforms is inherently tricky, Kuo says. He and coauthors used sequence information to map probes at both gene and exon levels, and applied spot quality filtering and other data preprocessing techniques to further improve consistency. Results indicate that microarrays perform more reliably than previously believed. Intra-platform consistencies are more than 99% for single-dye and 92%-99% for double-dye arrays for example.

"They've done a pretty good job of taking...