Samples genotyped using the Affymetrix 500 K chip set are available from the Affymetrix website [5]. Genotypes called using the DM algorithm are available for relatively large numbers of subjects. However preliminary studies revealed that this algorithm gave much higher rates of homozygosity, and longer and more frequent homozygous tracts, than the BRLMM algorithm [6]. Hence I decided to use only subjects for whom genotypes called using this latter algorithm were available. At the time the project was carried out these consisted of a small sample of ten subjects and their parents from ten different European (CEU) CEPH pedigrees whose genotypes could be downloaded from the website. (Subsequently, BRLMM genotypes have been made available for the whole HapMap sample.) The genotypes were used as provided from the site without any further quality control checks being applied.
In order to detect regions of homozygosity I looked for stretches in which there were at least ten contiguous markers which were homozygous. Interspersed markers with missing genotypes were ignored but the tract could not contain any markers called as heterozygous. The distance between the first and last marker had to reach 1 Mb or 5 Mb, according to the required tract length.
A number of different types of transmission error could be detected. For the purpose of the present analysis transmission errors were characterised on the premise that genotypes called assuming the presence of two alleles might actually reflect a different genotype if there were a cytogenetic abnormality, so that for example if a deletion had occurred then genotype A_ might be called as AA and if a trisomy were present genotype AAB might be called as AB. I identified "gained allele" errors when the child had an allele which neither parent possessed. However in order to look for transmission errors which might reflect the presence of either a deletion or uniparental isodisomy I counted as "lost allele" errors those in which one parent was AA, the other BB and the child either AA or BB.
I compared the observed number of "lost allele" errors occurring in regions of extended homozyogity to the number one would expect given the proportion of the markers falling within these regions in each subject. For simplicity, only autosomal chromosomes were studied. Markers on these chromosomes had an average spacing of 5.8 kb.