To further understand the workings of LRP, we randomly ordered the 35,528 typed individuals, and removed 50 of them at a time from the haplotype-sharing graph. Every time individuals were removed, we recomputed the fraction of individuals who had at least one surrogate parent (Fig. 2a), and the fraction of individuals belonging to the main, i.e. largest connected, cluster in the haplotype-sharing graph (Fig. 2b). The results were similar for the MHC region and the 10 Mb 15q25 region, and higher for the 15q25 6.4 Mb region. Specifically, with as little as 2% of the living population (~ 6300) typed, for the MHC and the 10 Mb 15q25 regions, about 78% of the individuals would have at least one surrogate parent and about 59% of the individuals belong to the main cluster. Similar results were achieved for the shorter 6.4 Mb 15q25 region by having about 1.5% (~ 4700) of the living population typed. With improved analytical tools (see below), and in some cases focusing on smaller regions, useful results could possibly be obtained with only 1% of the living
