February 28, 2011

FastIBD detection of identity by descent

I referred to this paper earlier in my post on the ERSA software. The authors write:
If the input data are phased, GERMLINE is between 2 and 3 orders of magnitude faster than ten runs of fastIBD. Phasing data with BEAGLE takes time that is similar to one run of fastIBD, so in practice, the computation time for ten runs of fastIBD is approximately one order of magnitude larger than the computation time for GERMLINE when the phasing step is included. However, the greatly improved accuracy of fastIBD compensates for the increased computing time.
I was looking at GERMLINE recently, but I think I will switch my attention to fastIBD. A few test runs have convinced me that it's got a reasonable running time on a run-of-the-mill PC, and the extra accuracy is probably worth the time.

The American Journal of Human Genetics, Volume 86, Issue 4, 526-539, 18 March 2010
doi:10.1016/j.ajhg.2010.02.021

High-Resolution Detection of Identity by Descent in Unrelated Individuals

Sharon R. Browning, and Brian L. Browning

Detection of recent identity by descent (IBD) in population samples is important for population-based linkage mapping and for highly accurate genotype imputation and haplotype-phase inference. We present a method for detection of recent IBD in population samples. Our method accounts for linkage disequilibrium between SNPs to enable full use of high-density SNP data. We find that our method can detect segments of a length of 2 cM with moderate power and negligible false discovery rate in Illumina 550K data in Northwestern Europeans. We compare our method with GERMLINE and PLINK, and we show that our method has a level of resolution that is significantly better than these existing methods, thus extending the usefulness of recent IBD in analysis of high-density SNP data. We survey four genomic regions in a sample of UK individuals of European descent and find that on average, at a given location, our method detects IBD in 2.7 per 10,000 pairs of individuals in Illumina 550K data. We also present methodology and results for detection of homozygosity by descent (HBD) and survey the whole genome in a sample of 1373 UK individuals of European descent. We detect HBD in 4.7 individuals per 10,000 on average at a given location. Our methodology is implemented in the freely available BEAGLE software package.

Link

1 comment:

Andrew Oh-Willeke said...

Is there a market for making identity by descent determinations faster?

The link between descent and appearance is often too attunated to be useful in criminal forensics, and in other circumstances (for example, using large data sets to reach historical population genetics conclusions), there is rarely a rush.