A multivariate prediction model for microarray cross-hybridization

Background: Expression microarray analysis is one of the most popular molecular diagnostictechniques in the post-genomic era. However, this technique faces the fundamental problem ofpotential cross-hybridization. This is a pervasive problem for both oligonucleotide and cDNAmicroarrays; it is considered particularly problematic for the latter. No comprehensive multivariatepredictive modeling has been performed to understand how multiple variables contribute to(cross-) hybridization.Results: We propose a systematic search strategy using multiple multivariate models [multiplelinear regressions, regression trees, and artificial neural network analyses (ANNs)] to select aneffective set of predictors for hybridization. We validate this approach on a set of DNAmicroarrays with cytochrome p450 family genes. The performance of our multiple multivariatemodels is compared with that of a recently proposed third-order polynomial regression methodthat uses percent identity as the sole predictor. All multivariate models agree that the 'mostcontiguous base pairs between probe and target sequences,' rather than percent identity, is thebest univariate predictor. The predictive power is improved by inclusion of additional nonlineareffects, in particular target GC content, when regression trees or ANNs are used.Conclusion: A systematic multivariate approach is provided to assess the importance of multiplesequence features for hybridization and of relationships among these features. This approach caneasily be applied to larger datasets. This will allow future developments of generalized hybridizationmodels that will be able to correct for false-positive cross-hybridization signals in expressionexperiments.