Evaluating the performance of oligonucleotide microarrays for bacterial strains with increasing genetic divergence from the reference strain

DNA oligonucleotide microarrays (oligoarrays) are being developed continuously; however, several issues regarding the applicability of these arrays for whole-genome DNA-DNA strain comparisons (genomotyping) have not been investigated. For example, the extent of false negatives (i.e., no hybridization signal is observed when the amino acid sequence is conserved but the nucleotide sequence has diverged to a level that does not allow hybridization) remains speculative. To provide quantitative answers to such questions, we performed competitive DNA-DNA oligoarray (60-mer) hybridizations with several fully sequenced (tester) strains and a reference strain (whose genome was used to design the oligoarray probes) of the genus Burkholderia and compared the experimental results obtained to the results predicted based on bioinformatic modeling of the probe-target pair using the available sequences. Our comparisons revealed that the fraction of the total probes that provided experimental results consistent with the predicted results decreased substantially with increasing divergence of the tester strain from the reference strain. The fractions were 90.8%, 84.3%, and 77.4% for tester strains showing 96% 89%, and 80% genome-aggregate average nucleotide identity (ANI) to the reference strain, respectively. New approaches to determine gene presence or absence based on the hybridization signal, which outperformed previous approaches (e.g., 92.9% accuracy versus 86.0% accuracy) and to normalize across different array experiments are also described. Collectively, our results suggest that the performance of oligoarrays is acceptable for tester strains showing >90% ANI to the reference strain and provide useful guidelines for using oligoarray applications in environmental gene detection and gene expression studies with strains other than the reference strain.