LOKI uses Markov chain Monte Carlo (MCMC) methodology to repeatedly sample possible segregation patterns. However, determining appropriate run length (number of iterations) is important for the accuracy of the IBD estimates. Using three separate sets of markers: (microsatellites, Affymetrix, and Illumina) we compared the average standard deviation of "phi2" (twice the kinship coefficient) between each pair of individuals in each pedigree at each centimorgan position on chromosome 7 for 10 replicates (from 10 different starting seeds) with 10,000, 100,000 and 1,000,000 iterations per replicate. To compare this information on different maps, we translated the genetic positions of the markers to the physical position based on NCBI build 34.3 (Figures 1 and 2). Ultimately, we used 1,000,000 iterations to compute IBD estimates for each White pedigree for the SNP maps. These computations were performed on a Beowulf-class computer cluster consisting of 60 dual processor nodes (25 dual Pentium II 350 MHz, 8 dual Pentium II 550 MHz, 18 dual Pentium III 800 MHz, 9 dual Pentium III 1,000 MHz), each with 512 MB of RAM; this provides an effective 18 GFLOP/s capacity (based on the Linpack benchmark [8]).
