Leiske Danielle L, Karimpour-Fard Anis, Hume Patrick S, Fairbanks Benjamin D, Gill Ryan T
Chemical and Biological Engineering, University of Colorado, Boulder, Colorado, USA.
BMC Genomics. 2006 Apr 4;7:72. doi: 10.1186/1471-2164-7-72.
DNA microarrays have proven powerful for functional genomics studies. Several technologies exist for the generation of whole-genome arrays. It is well documented that 25mer probes directed against different regions of the same gene produce variable signal intensity values. However, the extent to which this is true for probes of greater length (60mers) is not well characterized. Moreover, this information has not previously been reported for whole-genome arrays designed against bacteria, whose genomes may differ substantially in characteristics directly affecting microarray performance.
We report here an analysis of alternative 60mer probe designs for an in-situ synthesized oligonucleotide array for the GC rich, beta-proteobacterium Burkholderia cenocepacia. Probes were designed using the ArrayOligoSel3.5 software package and whole-genome microarrays synthesized by Agilent, Inc. using their in-situ, ink-jet technology platform. We first validated the quality of the microarrays as demonstrated by an average signal to noise ratio of >1000. Next, we determined that the variance of replicate probes (1178 total probes examined) of identical sequence was 3.8% whereas the variance of alternative probes (558 total alternative probes examined) designs was 9.5%. We determined that depending upon the definition, about 2.4% of replicate and 7.8% of alternative probes produced outlier conclusions. Finally, we determined none of the probe design subscores (GC content, internal repeat, binding energy and self annealment) produced by ArrayOligoSel3.5 were predictive or probes that produced outlier signals.
Our analysis demonstrated that the use of multiple probes per target sequence is not essential for in-situ synthesized 60mer oligonucleotide arrays designed against bacteria. Although probes producing outlier signals were identified, the use of ratios results in less than 10% of such outlier conclusions. We also determined that several different measures commonly utilized in probe design were not predictive of outlier probes.
