Microarray-based comparative genomic hybridization is leading to an increased understanding of bacterial evolution and patho-genesis by efficiently comparing whole genomes. Here, a sample genome is compared to a reference genome via comparative hybridization using two different fluorescent dyes. The logratio of the fluorescence intensities for sample and reference genome for each gene on the array is then usually compared to a cutoff classifying the belonging gene as absent or present with respect to the sample genome. The resulting list of candidate absent genes then undergoes confirmational analyses, i.e. PCR or sequencing, which are decisive with respect to both time and costs of the experiment. Thus, there is vital interest to reduce the rate of false positives in the list of candidate absent genes from the comparative genomic hybridization step. Our approach to accomplish this task uses the expected relationship between the logratio and the mean log intensities for absent genes to linear transform the inten-sity data before comparing to a cutoff. For validated data from a series of comparative genomic hybridizations for two Bartonella species we assessed significance and efficacy of the new approach. We show that we are able to halve the rate of false positives in the list of candidate absent genes for a typical comparative genomic hybridization experiment, thus significantly reducing time and costs of necessary confirmational analyses.