Homogeneous probe hybridization is a prerequisite for the robust design of microarrays. Elaborate algorithms were developed to select for probe sets with uniform melting temperatures (Tm). However, at least short oligonucleotides (<20 bp) show large variation in the on-chip hybridization efficiency even if designed with state-of-the-art algorithms. This variation can be explained by steric effects and interferences on the solid surface as well as by chemical conditions that may deviate from conditions used to develop the algorithms.
We designed 412 random 13mer duplexes to study the differences between the Tm of nearest-neighbor algorithms and the Tm values measured by dissociation analysis in polymerase chain reaction (PCR) and 1X SSC buffer. We tested the effects of theoretical vs empirical Tm values on the hybridization variation of 40 duplexes on-chip. Although the empirical approach resulted in a slightly better prediction of hybridization efficiency, less than one-fifth (17%) of the observed variation could be explained by the factor Tm alone. We conclude that state-of-the-art algorithms can be used for a first selection of short oligonucleotide probes, but that it is then necessary to perform an on-chip selection to obtain a probe set with a uniform hybridization behavior.