Juxtapositions of short IBD blocks can cause biased estimation in inferences based on the length of IBD blocks. C. W. K. Chiang1, J. Novembre2 1) Ecology and Evolutionary Biology, UCLA, Los Angeles, CA; 2) Human Genetics, University of Chicago, Chicago, IL.
Blocks of identity-by-descent (IBD) play an important role in many modern genetic applications, including long-range phasing, imputation, genetic mapping, detection of natural selection, and demographic inferences. One commonly used definition of IBD blocks is that they are contiguous segments of the genome inherited from a recent shared common ancestor without intervening recombination. With programs like Beagles fastibd, long IBD blocks (>1cM) can be efficiently detected using high-density SNP array data of a population sample. However, all programs detect IBD based on contiguous segments of identity-by-state (IBS). As such, detected IBD blocks could often be due to the juxtaposition of smaller IBD blocks inherited from different common ancestors. Here, we show the juxtaposition of small IBD blocks leads to an error in estimating the length distribution of IBD blocks and can affect downstream inferences. To demonstrate the prevalence of subsegment juxtapositions, we used coalescent simulations where we know the precise genealogy of the sample and found that >35% of the detected IBD segments of 1cM or longer are composed of at least two subsegments. In particular, 11% of the detectable segments consist of at least 1 other subsegment >25% of the total length, and this effect was more pronounced for detectable segments between 1 to 2cM long, compared to segments >2cM long. To demonstrate that the juxtaposition can lead to practical problems, we investigated the impact on a novel estimator of the de novo mutation rate using IBD blocks. We observed accurate estimates of the input mutation rate when true IBD blocks are used, but overestimates of the mutation rate by ~15 fold using inferred IBD blocks. When the effect of juxtaposition on the estimated age of the block was modeled, the mutation rate estimate improved greatly. Our results suggest that identifying IBD blocks based on extended IBS can inflate the length of IBD blocks, and in this case results in an inflated estimate of the de novo mutation rate, unless properly accounted for. This effect should be carefully considered as methods to detect shorter IBD blocks using sequencing data are being developed.
You may contact the first author (during and after the meeting) at