The Effect of SNP Marker Density on the Efficacy of Haplotype Tagging SNPs – a Warning

• Published in: Annals of human genetics 2005-03, Vol.69 (2), p.209-215
• Main Author: Iles, Mark M.
• Format: Article
• Language: English
• Subjects: Chromosomes, Human, Pair 20; Haplotypes; Humans; Polymorphism, Single Nucleotide
• ISSN: 0003-4800; 1469-1809
• DOI: 10.1046/j.1469-1809.2004.00141.x

SUMMARY We investigate here the efficacy of selecting haplotype tagging SNPs at different marker densities (2kb‐10kb). Our results are based on publicly available data on 5324 markers with a median spacing of 1kb from chromosome 20. We find that whatever density of SNPs is used, htSNP analysis indicates in most cases that at least 80% of the variation can be captured using a subset of SNPs. However, as marker density decreases these htSNPs become increasingly unreliable. In this dataset htSNPs were selected to capture at least 80% of the variation at every observed SNP. At an observed SNP density of 2kb, htSNP analysis suggests that the htSNPs capture on average 95% of the observed variation, when in fact they capture 88% of the unobserved variation. At a density of 10kb, htSNP analysis suggests that 93% of the observed variation was captured, when in fact they capture on average only 78%. Our results indicate that htSNP analysis is only reliable when markers are dense – a spacing of even 2kb shows a considerable loss of information. Such findings are important both for individual studies utilising htSNPs to reduce costs, and for projects such as HapMap which try to characterise human genomic variation using htSNPs.