Also, statistical probabilities for unique matches and assumptions about the database population have recently come into question. Arizona compared felons already in its database and reported 20 "hits" between offenders where both alleles matched at 10 loci. Only three would be expected based on established calculations. The extras could be relatives or even entirely coincidental matches.

Bieber cited two high-profile cases in which investigators used partial matches to narrow in on a perpetrator who was a brother or uncle. But unless the shared alleles are quite rare or the matches are extremely close, this approach can be very inefficient. So Bieber suggested the type of kinship analysis and weighted allele probabilities used to connect family members with loved ones lost in the World Trade Center attacks. CODIS software does not allow for such sophisticated searches at the moment, but Bieber predicts an upgrade as soon as investigators see the potential and define clear statistical thresholds that must be met.

Comments