Machine Learning Applications

Bonaparte Disaster Victim Identification System

Using modern DNA technology, victim identification has become much more reliable as victims can be identified by comparing the extracted DNA with that of a reported missing person, or, absent such DNA, with that of possible relatives.

In mass fatality incidents, it is infeasible to carry out DNA matching by hand as there are too many combinations to compare. Consider a case with 10 victims with their 10 putative pedigrees. This results in just 100 combinations, but 100 victims with their 100 pedigrees yields 10,000 combinations. And then these samples have not even been checked against each other (4,950 combinations) or checked for contamination (100 × the number of elimination profiles). An automated routine is then indispensable.

The Bonaparte system has been developed as a tool for forensic researchers to automatically carry out DNA matching. Identification is achieved by using probabilistic methods to model Mendelian inheritance. In particular, Bayesian networks are very well suited to model the statistical relations of genetic material of relatives in a pedigree. They can directly be applied in kinship analysis with any type of pedigree of relatives of the missing persons. An additional advantage of a Bayesian network approach is that it makes the analysis tool more transparent and flexible, allowing to incorporate other factors that play a role – such as measurement error probability, missing data, statistics of more advanced genetic markers etc.

The development of Bonaparte is done for and in collaboration with the Netherlands Forensic Institute (NFI). In 2010, NFI has succesfully deployed Bonaparte to identify victims from the Afriqiyah Airways crash in Tripoli, Libya.