NCJ Number
203497
Journal
Journal of Forensic Sciences Volume: 48 Issue: 6 Dated: November 2003 Pages: 1239-1248
Date Published
November 2003
Length
10 pages
Annotation
This article discusses a measure of estimating the relatedness of two individuals and an algorithm that uses this measure to cluster individuals.
Abstract
DNA markers are routinely used to determine the correct familial relationship between persons or animals. This study involved the problem of determining correct family relations in a large group of individuals, focusing on the case where no nongenetic information was available. It is assumed that the true pedigree linking the individuals consists of many smaller sub-pedigrees with very distant relationships between the sub-pedigrees. The study defines a distance measure between two individuals in a pedigree, shows how to estimate the distance between a pair of individuals, and gives an algorithm that generates a partition of individuals into groups using the estimated distances. The effectiveness of this method is illustrated on some examples and in a simulation study. These examples and simulations suggest that for identifying very close relationships, the method is quite reliable with error being quite rare, a desirable feature of a clustering algorithm. The use of the clustering algorithm in a two-stage pedigree reconstruction procedure is discussed. Directions in which the analysis could be extended include performing a Bayesian estimation procedure in which priors are put onto the various distances so that the whole posterior distribution of distances as a measure of the relationship between two persons may be found. The estimated distances could be used in a pedigree search to restrict the pedigrees to look at. Applications include the identification of family groups among bodies found in mass graves and identification of family groups in animal populations. 9 tables, 9 figures, 7 references