Tag SNP Selection in Genotype Data for Maximizing SNP Prediction Accuracy
Gad Kimmel, Tel Aviv University

The search for genetic regions associated with complex disease, such as cancer or Alzheimer's disease, is an important challenge that may lead to better diagnosis and treatment. The existence of millions of DNA variations, primarily single nucleotide polymorphisms (SNPs), may allow the fine dissection of such associations. However, studies seeking disease association are limited by the cost of genotyping SNPs. Therefore, it is essential to find a small subset of informative SNPs (tag SNPs) that may be used as good representatives of the rest of the SNPs.

We define a new natural measure for evaluating the prediction accuracy of a set of tag SNPs, and use it to develop a new method for tag SNPs selection. Our method is based on a novel algorithm that predicts the values of the rest of the SNPs given the tag SNPs. In contrast to most previous methods, our prediction algorithm uses the genotype information and not the haplotype information of the tag SNPs. Our method is very efficient, and it does not rely on having a block partition of the genomic region.

We compared our method to two state of the art tag SNP selection algorithms on 58 different genotype data sets from four different sources. Our method consistently found tag SNPs with considerably better prediction ability than the other methods.

Joint work with Eran Halperin and Ron Shamir.