NCJ Number
219257
Journal
Journal of Forensic Sciences Volume: 52 Issue: 4 Dated: July 2007 Pages: 895-897
Date Published
July 2007
Length
3 pages
Annotation
This paper reports on the adaptation of two Bayesian networks (BNs) to a daughter-grandmother relationship and a half-sister/half-sister relationship in resolving paternity relationships with X-chromosome short tandem repeat (STR) genotyping.
Abstract
In the first case, the BN yielded a paternity probability of 99.98 percent, which was sufficiently high to support the legal claim of paternity in France, where the case was processed. In the second case, the BN found paternity probabilities of 99.98 percent for one half-sister and 99.99 percent for the second half-sister as having the same father. In both cases, the use of BNs was critical in enhancing the likelihood ratio for paternity. In the first case, a woman (AM) claimed to be the daughter of MG and a deceased man (JK). The parents of JK contested the alleged paternity. In the second case, two alleged half-sisters suspected common paternity. A BN is a graphic system of nodes and directional arrows that is constructed by the analyst in modeling a set of random variables. The model portrays the possible states of each variable as well as a conditional probability table. The conditional probabilities of a node depend on the states of its "parent" nodes (the ones from which directional arrows arise). When applied to kinship studies that use STRs, a specific BN can be modeled upon a specific pedigree. The BN describes alleged blood relationships, specific rules of inheritance of alleles (genes). Rules can include various inheritance schemes such as autosomal, patrilineal (Y-chromosomal), X chromosomal, and matrilineal (mitochondrial). Exceptions to the rules (mutations, null alleles, dropout, etc.) can be included as well. Allelic types/locus are entered as the BN’s data for the case. The BN considers alternative hypotheses and a likelihood ratio are computed for all the data. 3 figures and 8 references