Abstract
A paper by Zhang et al. in the July 2003 issue of Nature Biotechnology
presents PerfectMatch, a physical-model-based algorithm for modeling
the binding interactions between probes on Affymetrix (Santa Clara,
CA, USA) GeneChip oligonucleotide arrays and their target nucleic
acids. Stochastic-model-based algorithms have been shown to improve
the preprocessing of array data to take into account the presence
of optical noise, nonspecific binding and probe-specific effects.
Although Zhang et al. demonstrate the superior performance of PerfectMatch
over two widely used algorithms, MAS 5.0 and dChip, they fail to
consider a leading stochastic-model-based procedure, robust multi-array
analysis (RMA), from our group2. Here, we provide evidence that (i)
measurements obtained using RMA are as precise and more accurate
than those obtained using PerfectMatch and (ii) the accuracy of stochastic-model-based
procedures can be improved by carefully incorporating physical model
information.
Users
Please
log in to take part in the discussion (add own reviews or comments).