Until recently, the DNA profiles that have been generated for forensic purposes have been almost exclusively those that could be characterized as “probable cause matches,” in which DNA testing has been performed upon a reference sample taken from a suspect that has already been linked to a crime by direct or circumstantial evidence. A new category of DNA profile “matches” are becoming increasingly common however – those that are generated as a result of “cold hits” that result from the trawling of a large number of DNA profiles maintained in databases (usually those of previously convicted offenders). Since the primary difference between these kinds of matches is the manner in which a suspect is first identified, it is generally accepted that it is not possible to convert one type of case into the other (for instance, by simply retesting a reference sample once a “cold hit” has been identified). It is also generally accepted in the scientific community that the statistical significance of those two kinds of DNA profile matches should be determined differently. However, there are at least three different commonly held opinions on how the statistics associated with “cold hits” should be generated and presented.
The first group to address this issue was a body of experts appointed to the Committee on DNA Science by the National Research Council in 1992. The position of this group is that database searches should be used to identify potential suspects but not to calculate frequency estimates. When successful, suspects identified by these searches would then be tested at a completely different group of independent genetic markers that would also be compared to the evidence. If these additional genetic loci also match between the suspect and evidence sample, they alone would be used to compute probabilities that reflect the significance of a match. With this methodology the genetic markers used in the original database search are specifically and deliberately excluded from any statistical calculation.
A second committee of prominent experts advocated a significantly different approach in 1996. They specifically recommended that, “When the suspect is found by a search of DNA databases, the randommatch probability should be multiplied by N, the number of persons in the database.” (The Evaluation of Forensic DNA Evidence, 1996, National Research Council Press, p. 40, 161). Proponents of this approach feel that the first method is too conservative. Their alternative method differs in three ways: 1) no testing is performed at additional loci; 2) genetic markers used in the original database search are included in the statistical calculations; and 3) the size of the database being searched (N) is taken into consideration.
A third group is comprised of individual scientists who have published peer-reviewed manuscripts in which they argue that a “cold hit” should actually be given more weight than a match found in a “probable cause” case. Their position is based on the thinking that not only has the defendant been found to match the evidence, but many more individuals have been found to not match. In “probable cause” cases where only a single match is found during the course of DNA testing, there is at least still a formal possibility that one or more untested people may also match the evidence –that possibility becomes increasingly less likely as the database used for a cold hit becomes larger. Proponents of this approach also feel that the first method is too conservative. Their method differs from it in three ways: 1) no testing is performed at additional loci; 2) genetic markers used in the original database search are included in the statistical calculations; and 3) the size of the database being searched (N) is taken into consideration. It also differs from the second in one very important way: the effect of the database size on the significance of a match is precisely opposite – large databases generate the most damning statistics for a defendant while, in the second approach, the larger the database the less damning the statistics become to a defendant. The second and third approaches are diametrically opposed with respect to implications of the size of the database that is searched.
The proponents of each of these three approaches include many eminent scholars in the field of genetics and statistics. For instance, the blue ribbon panel of experts that generated the first National Research Council on DNA typing report (which supports the first approach as described in paragraph 4 above) includes Drs. Mary-Claire King, Richard Lempert, Eric Lander, Ruth Macklin, Thomass Marr, Victor McKusick, Philip Reilly and Sandy Zabel. Members of the second National Research Council on DNA Typing (which recommends the second approach as described in paragraph 5 above) include prominent population geneticists and statisticians such as Drs. James Crow, Arno Motulsky, Thomas Nagylaki, Mashotoshi Nei, David Siegmond and Stephen Stigler. The third approach (described in paragraph 6 above) is one that has been principally advocated by very influential and often cited geneticists and statisticians such as Drs. David Balding, Peter Donnelly and Bruce Weir (as in publications such as: Errors and Misunderstandings in the Second NRC Report, D. J. Balding, Jurimetrics, Summer 1997, 37:469-476; Evaluating DNA Profile Evidence When the Suspect s Identified through a Database Search, D. J. Balding and P. Donnelly, Journal of Forensic Science, 1996, 41:603-607; and Interpreting DNA Evidence, I. W. Evett and B. S. Weir, Sinauer Press, 1998, pp. 219- 222). This appears to represent a genuine split between three fundamentally different approaches by experts who are significant both in number and in eminence within their fields.
Hille, Karl B. (2003). 'Cold hit' DNA tests get mixed reviews. The Winchester Star. July 29, 2003.
Krane, D. (2002). Declaration. People of the state of California vs. Paul Robinson.
Mueller, L. (2002). Declaration. People of the state of California vs. Paul Robinson.