SNPs on Chips: The hidden genetic code in expression arrays

Margit Burmeister; Fan Meng; Elzbieta Sliwerska; Terence P. Speed; Edward G. Jones; William E. Bunney; Huda Akil. Stanley J. Watson
XIIIth World Congress of Psychiatric Genetics. 2005.


Gene expression microarray analysis is one of the fastest growing fields in psychiatric research. In contrast, genetic studies identify SNPs associated with these disorders. Combining these ideas, we determined both gene expression and >50 SNP genotypes in >60 brain samples from subjects affected with affective disorders or schizophrenia and controls. The common Val(108/158)Met SNP in COMT was found significantly associated with expression, consistent in >10 brain regions. Increased expression seemed to compensate for the lower enzymatic activity of the Met allele. However, further analysis showed that this "expression" difference was driven by the presence of the SNP itself in the central region of one of the COMT probes on the Affymetrix array. Once taken into account, no significant expression difference remained. This artifact may be quite common: The Affymetrix 133AB Plus chip contains >50,000 SNPs, with >5,000 SNPs that are both common and in the central probe region, affecting hybridization. Conversely, genotype information on these SNPs may be extractable from expression data, by comparing the signal from an allele-specific probe to the overall expression level of the gene measured by other probes. Applying this idea to our dataset, we predicted that one SNP would differ significantly in allele frequency between bipolars and controls. Genotyping the 25 samples on which expression data were available confirmed that allele frequencies differed significantly (p=0.015). While this difference disappeared once more samples were genotyped, we conclude that SNPs hidden in microarray experiments can not only misguide expression studies, but may also become useful to identify candidate SNPs for psychiatric disorders.