Updated

Genetic profiles may not be as reliable in pinpointing culprits as the FBI would have investigators — and juries — believe.

According to a Los Angeles Times investigation published Sunday, the odds for finding genetic similarities between unrelated people are higher than commonly estimated.

Arizona state crime lab analyst Kathryn Troyer found dozens of remarkably similar genetic matches by hunting through the FBI's Arizona DNA database, leading defense attorneys to request similar searches in an effort to exonerate their clients.

Saying Troyer's findings are misleading and the searches could violate privacy laws, the FBI has sought to block similar hunts.

Genetic evidence has so far proven all but irrefutable in linking suspects to crime scenes. The FBI puts the odds of one person's genetic profile matching another's as high as 1 in 100 billion. But no one knows with any precision how rare DNA profiles are.

Each person's genetic makeup is unique, but their genetic profile — a sliver of their entire genome — is not as unique. Siblings often share most of the same genetic markers — or loci, locations on the chromosome that are commonly used to identify people — but not all. States usually look at 13 loci in a DNA sample.

Troyer found dozens of samples in which two unrelated men shared nine or 10 loci, amounts commonly seen among relatives.

Defense attorneys have requested judges grant them similar searches for comparison in other state databases, but the FBI has successfully argued against the move in California and other states, saying databases are to be used only by criminal justice agencies.

But in two states, judges overruled the agency's objections. The subsequent searches yielded nearly 1,000 pairs that matched at nine or more loci.

FBI officials say that the similarities were to be expected given the type of searches employed. Moreover, the non-matching loci would be used to rule out a suspect and thus ensure the wrong person is not implicated. Some of the cases of matching loci could also mean relatives are in the databases.