Genetic maps generated from the 1000G phased OMNI data. 

OMNI haplotypes were obtained from the Phase 1 dataset 
(/vol1/ftp/phase1/analysis_results/supporting/omni_haplotypes/).

Genetic maps were generated for each population separately using LDhat 
(http://ldhat.sourceforge.net/). Haplotypes were split into 2000 SNP windows
with an overlap of 200 SNPs between each window. The recombination rate was
estimated for each window independently, using a block penalty of 5 for a 
total of 22.5 million iterations with a sample being taken from the MCMC 
chain every 15,000 iterations. The first 7.5 million iterations were 
discarded as burn in. Once rates were estimated, windows were merged by 
splicing the estimates at the mid-point of the overlapping regions.

LDhat estimates the population genetic recombination rate, rho = 4Ner. In 
order to convert to per-generation rates (measured in cM/Mb), the LDhat
rates were compared to pedigree-based rates from Kong et al. (2010). 
Specifically, rates were binned at the 5Mb scale, and a linear regression
performed between the two datasets. The gradient of this line gives an 
estimate of 4Ne, allowing the population based rates to be converted to 
per-generation rates.

Using the above method, the effective population sizes for each population
were estimated as follows.

ACB	17939.993963
ASW	17150.137312
CDX	12255.652372
CEU	11490.672741
CHB	13038.180071
CHS	11962.133624
CLM	10862.696703
FIN	9648.177725
GBR	11314.802435
GIH	12990.359505
IBS	11635.830405
JPT	11600.019740
KHV	12339.783775
LWK	18861.336542
MKK	16951.674498
MXL	9995.936127
PEL	8315.793426
PUR	11946.510129
TSI	12787.027208
YRI	15898.782067


Adam Auton (adam.auton@einstein.yu.edu)
May 7th, 2013