Sandra M. Sovinski, B.S., Patricia P. Hamby, M.S., Gabriel E. Novick, M.D., Ph.D. and Mohammed A. Tahir, Ph.D.
Indianapolis-Marion County Forensic Services Agency, 40 South Alabama Street, Indianapolis, Indiana 46204

× Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø × Ø

In comparing two samples, what if there was a match at D1S7 with a frequency of 1 in 79? Not enough? Add another matching locus, D2S44, for a combined frequency of 1 in 8,200. Still not enough? Add another, D10S28, for a frequency of 1 in 1,200,00. Still not satisfied? Add D17S79 and get a combined four-probe match frequency of 1 in 10,000,000. Who wouldn’t be convinced?

We report our discovery of such a match between blood samples from two different individuals. Two database samples matched at four loci, D1S7, D2S44, D10S28 and D17S79, appearing at first to be duplicates. Upon developing a fifth probe, D4S139, and finding that the samples were not duplicates, we knew then that we had siblings, since the only other alternative was that the significance of a four-probe match was not at all what the entire scientific community thought it was!

Our database samples were collected in part from police officers, so we were able to obtain additional blood samples from the two donor brothers. Since our database records did not include names, we relied on birthdate information. The new samples were run along with the original database samples to confirm that they were indeed the same donors.

Had this been a case match, with the wrong brother accused of the crime and with results from only four probes, after applying the formula for unrelated individuals, 2p_ip_j and the product rule, the estimated frequency of the match would have been reported at 1 in 10,000,000 for the Caucasian population. Recalculating the estimated frequency of such a match between brothers, we again applied the product rule but to an appropriate formula, (1 + p_i + p_j + 2p_ip_j)/4, and obtained a frequency of 1 in 102. Obviously there is a huge difference between the two estimates, as one would expect. But, with the odds at 1 in 100, we were still surprised that we found two siblings who matched at four probes.

An additional consideration is the RFLP population database samples themselves. Since these samples provide the basis for statistical estimation of the frequency of a pattern in unrelated individuals, it is important that samples are collected as randomly as possible. The presence of a duplicate sample directly affects the resulting frequency data, as does the presence of related individuals. Throughout conservative methods of estimating match frequency however, such effects on final data should be negligible.