Model Based Probe Fitting and Selection for SNP Array

Recent advances of high-throughput SNP arrays such as Affymetrix’s GeneChip Human Mapping 500K array set have made it possible to genotype large samples in a fast and cheap manner. A lot of algorithms were developed to call the genotypes from SNP array. When considering the low level preprocessing of SNP array, most algorithms just borrow the techniques from the gene expression microarray. As in the analysis of gene expression microarray, the low level preprocessing of SNP array, e.g. probe summarization, is very important in the analysis of SNP array such as genotyping, loss of heterozygous (LOH) inference, and copy number inference. In this paper, we present a model based method for probe fitting and selection of SNP array. This method exploits the abundant high quality genotypes such as HapMap data, which were genotyped and validated by several independent genotyping techniques. The new probe summarization method can be used with the existing genotyping methods to improve the accuracy.