Motivation: Recently, much attention has focused on using prediction from population genetic theory to quantify variation in recombination rate along the human genome owing to the promise of association or linkage disequilibrium (LD) mapping to identify genes underlying complex traits. Current state of the art approaches to the problem estimate the local population recombination rate from patterns of LD among common single nucleotide polymorphisms (SNPs) assuming the population is randomly mating and constant in size. Results: Here we describe an alternative method that can accommodate complex population structure and ascertainment bias. Using multiple linear regression and non-parametric bootstrap re-sampling, our method uses the variances and co-variances of un-phased SNPs at different frequencies to estimate the local recombination rate. We evaluate this new approach via Monte Carlo simulation and compare its performance with three other available methods. Our approach is less biased when the demographic assumptions of the standard neutral model are violated. We also apply our approach to the well-characterized hot spots near the human TAP2 gene and a 206-kb region on human chromosome 1q42.3 near minisatellite MS32. The results are consistent with findings in literatures.
展开▼
