Sources: Joel Rubin and Maura Dolan, “DNA Search Fails to Find Relatives of Unknown Serial Killer,” Los Angeles Times, December 3, 2008; Maura Dolan, Joel Rubin, and Mitchell Landsberg, “DNA Leads to Arrest in Grim Sleeper Killings,” Los Angeles Times, July 8, 2010.
More Advanced Familial Searching Techniques
Population geneticists and biostatisticians have recommended using more advanced statistical techniques for conducting familial searches as a way of improving both efficiency and accuracy. The most commonly agreed-upon approach is to use “likelihood ratios” (LRs). As applied to familial searching, this is a statistical method that goes beyond simple allele counting and instead evaluates the genetic evidence to support the likelihood that two individuals are related compared with the likelihood that they are not. In general, the LR is the ratio of two probabilities of the same event under different hypotheses. For DNA testing the LR is the probability of the observed genetic profiles given proposed familial relationships versus the probability of observing the genetic profiles if the donor of the evidence and the identified partially matching profile sources are unrelated.21 Charles Brenner, a consultant on paternity testing and forensic DNA practices, has described an LR as follows: “[It is the] amount of coincidence that you have to swallow if you want to believe that two profiles are similar merely by chance.”22
The benefit of using LRs is that they make better use of genetic information and produce a prioritized list of partial matches. In addition to comparing the amount of sharing between any two individuals with the amount of sharing that would be expected if those individuals were siblings or parent and offspring, the LR can take into account the relative frequencies of the matching alleles in the population. The rarer the alleles that match, the more likely it is that the match indicates a potential familial relationship.
Frederick Bieber and colleagues performed some mathematical simulations to investigate the chances of finding a true relative using this approach. Specifically, they ran a series of simulations by comparing an “unknown” sample with each registered offender in the database to determine the likelihood ratios of successfully identifying biological relatives of the offenders. Assuming a database size of 50,000, they argue that a parent of a child of a known offender would be identified 62 percent of the time as the very first lead (highest likelihood ratio) and 99 percent of the time among the first 100 leads (99 families would be investigated to find the one family related to the perpetrator). This assumes that the relative is in the database in the first place, and that the crime is committed in the state where that relative resides.23 They conclude: “Our simulations demonstrate that kinship analysis would be valuable now for detecting potential suspects who are the parents, children, or siblings of those whose profiles are in forensic databases.”24 LRs can also weigh the significance of a partial match on the basis of the racial and ethnic populations from which the DNA samples were drawn. For example, in the Caucasian population fathers and sons share an average of 16 out of the 26 CODIS-designated alleles, while two completely unrelated Caucasian individuals will share an average of 9 alleles. These figures will vary slightly across racial and ethnic populations. In groups that are more isolated and where there has been more inbreeding (e.g., cousins having children with one another), the number of shared alleles will be greater, even for so-called unrelated individuals. On the other hand, populations where there has been a lot of breeding across races and cultures and that have exhibited more mobility will show more allelic diversity and thus less allele sharing.
Although Bieber and colleagues are optimistic about the value of conducting familial searches using current techniques, others have pointed out that even these more refined techniques will produce a high number of false positives, especially when used in the context of large databases. High numbers of false leads result in resource-intensive investigations. In addition, these techniques are not likely to turn up any significant investigatory leads beyond full siblings and parental relationships.25
One way to narrow potentially long lists of possible relatives of suspects is to subject the stored DNA samples identified in a partial match to additional genetic testing. Kristen Lewis and colleagues advocate the use of ancestrytesting techniques—specifically Y-chromosome analysis and mitochondrial DNA sequencing—to exclude individuals from investigation who cannot be related to the true perpetrator through either paternal or maternal lineage.26 Y-chromosome (Y-STR) analysis involves examining genetic markers along the Y chromosome; since the Y chromosome is paternally inherited, it can be used to trace family relationships among males. Similarly, mitochondrial DNA (mtDNA) is inherited from the mother and can be used to trace back maternal lineages. Frederick Bieber and David Lazer suggest that Y-STR analysis alone could eliminate 99 percent of false leads.27 Others have recommended expanding the scope of genetic markers used in the DNA profile as a way of narrowing long lists of potential relatives. All these techniques would require accessing the stored biological samples of individuals in the database and subjecting them to additional genetic tests.
U.S. Policy on Familial Searches
In the United States the practice of familial searching was limited by a policy adopted by the FBI that prohibited the release of any identifying information about an offender in one state’s database to officials in another state unless the offender’s DNA was an exact match with the DNA evidence found at the scene of the crime (with limited exceptions for cases where slight differences between the profiles could be explained by degradation). In the summer of 2006 the FBI changed its policy in response to a request from Denver authorities after a close match was found between evidence taken from the scene of a rape and a convicted felon in Oregon, indicating that he was a potential relative of the actual perpetrator.28 The interim policy, which became effective on July 14, 2006, allows states to share information related to “partial matches” upon FBI approval.29 Following FBI authorization, discretion whether to release the offender information is left to the state that holds the information. According to the FBI memorandum distributed on July 20, 2006, “For situations in which there is no other available investigative information, the FBI Laboratory has instituted an Interim Plan that may permit the release of the offender’s identifying information.”30
The FBI’s interim plan defines a “partial match” as “a moderate stringency candidate match between two single-source profiles having at each locus at least one allele in common.”31 This plan did not allow information to be released between states as a result of low-stringency searches. In this way the FBI was insisting that the search parameters initially established for CODIS remain intact. Furthermore, the FBI distinguishes these partial matches from familial searching, claiming that these partial matches are “inadvertent,” as opposed to purposeful searches for incomplete matches. According to Tom Callaghan, FBI CODIS unit director, “The FBI does not do familial searching.”32
The FBI’s distinction between an “inadvertent partial match” and familial searching seems tenuous; after all, the follow-up to such a partial match is to seek out family members of that individual. Nonetheless, the FBI’s insistence that partial-match follow-ups do not constitute familial searching may indicate the agency’s own recognition that full-scale trawling of the databases is not authorized under current law and that allowing this more aggressive searching would render the database operations vulnerable to legal challenge. Whether legislative authority would be needed to authorize familial searching was a central question posed by the FBI itself at its national symposium on familial searching in March 2008.33
To be fair, the DNA Identification Act of 1994 that established CODIS is silent on the issue of familial searching and partial matches and simply states that the database will be used “for law enforcement purposes.”34 Nonetheless, there is some indication in the legislative history that Congress certainly did not intend for the database to be used to trawl for individuals other than those in the database. For example, Senator Herb
Kohl from Wisconsin stated on the Senate floor in support of the DNA Analysis and Backlog Elimination Act of 2000:
Currently, all 50 states require DNA samples to be obtained from certain convicted offenders, and these samples increasingly can be shared through a national DNA database established by Federal law. This national database . . . enables law enforcement officials to link DNA evidence found at a crime scene with any suspect whose DNA is already on file. By identifying repeat offenders, this system does make a difference.35
In other words, the idea was to create a database of known, convicted offenders so that law-enforcement officials could link those offenders to other crimes they might have committed and to have a way of catching them if they acted again. It was not to use the database as an intelligence or surveillance tool to investigate other people who were not in the database.
Perhaps the more significant evidence that familial searching crosses a well-established line is in the design of CODIS itself. Familial searching of the database is not a new concept; it was well known during the time at which CODIS was established that the database could be used in this fashion. In 1992, two years before the establishment of CODIS, the National Academy of Sciences issued a comprehensive report that made a series of recommendations for DNA data banking and testing. The report addressed head-on the issue of familial searching:
The ability of DNA to recognize relatedness poses a novel privacy issue for DNA databanks. . . . To put it succinctly, DNA databanks have the ability to point not just to individuals but to entire families—including relatives who have committed no crime. Clearly, this poses serious issues of privacy and fairness. . . . It is inappropriate, for reasons of privacy, to search databanks of DNA from convicted criminals in such a fashion. Such uses should be prevented both by limitations on the software for search and by statutory guarantees of privacy.36
In addition, the Privacy Act of 1974 requires that a formal rule making be issued for any new government database that defines clearly the categories of individuals who will be affected by the database.37 Such rules were issued by the U.S. Department of Justice for the National DNA Index System (NDIS) at its inception in 1998. Those rules established four categories of individuals covered by the system: convicted offenders; missing persons and their close biological relatives; victims; and DNA personnel. In addition, the rules included the following safeguard: “NDIS will disclose to a criminal justice agency the DNA records of another criminal justice agency only when there is a potential DNA match.”38 Given that family members represent an entirely separate category of individuals affected by the database system, and that partial matches result in disclosure of information between agencies when a DNA match is not complete, it seems that, at the very least, a formal rulemaking process is required under the Privacy Act in order for the database to be used to identify potential family members.
The agency’s hesitation to cross the line was at one time shared by most state database administrators. At the national symposium on familial searching in March 2008, sentiments were nearly unanimous among the 50 state CODIS administrators and their legal representatives that they did not currently have authority to move forward with familial searching.39
Even so, it is not clear that the line the FBI is attempting to draw between “partial matches” and “familial searching” will hold as a practical matter. A memorandum dated August 2, 2007, from Tom Callaghan stated that a “Next Generation CODIS” was under development that would provide a new search engine capable of performing “joint pedigree likelihood ratio” analyses.40 Although the stated intent of the upgrade is to conduct missing person searches, it is hard to imagine how the states or the FBI will be prevented from using this for familial searches in criminal investigations. In the meantime, the interim policy appears to have encouraged some familial searching advocates to push harder at the state level to initiate statewide familial searching programs. Several states, including Florida, South Carolina, North Carolina, Colorado, Missouri, Oregon, Arizona, and Massachusetts, have agreed to disclose partial matches to law-enforcement agencies in accordance with the FBI interim policy. These states have generally focused almost exclusively on the technicalities of familial searching and very little on the political, legal, or ethical ramifications of such searches. In contrast, Maryland statutorily banned the use of familial DNA searches.41
BOX 4.4 A 14-Year-Old Offender’s DNA Leads Police to a Murder Suspect
A 20-year-old woman named Lynette White was fatally stabbed in southern Wales in 1988. It was one of the most brutal murders in Welsh history. The case went cold for 12 years, but the police were not ready to give it up. In 2000 forensic investigators completed a new DNA sweep of the victim’s apartment, hoping to acquire new forensic evidence. The sweep turned up spots of blood on a baseboard that had been missed on the first search. They profiled the bloodstain DNA and compared it with profiles in their national database but did not find an exact match. However, an allele in the crime-scene DNA was found in only 1 to 2 percent of the profiles in the NDNAD. By using a low-stringency familial search combined with geographical constraints, police found approximately 70 potential relatives of the person who had left the crime-scene stain. Forensic investigators eventually found a person in the NDNAD who had a reasonably close DNA profile to the crime-scene DNA—a 14-year-old boy who was not alive when Lynette White was murdered. The boy’s DNA was in the database because he had previously gotten into trouble with the police. The police began looking into the boy’s family and focused attention on the boy’s paternal uncle, Jeffrey Gafoor. Gafoor’s DNA was an exact match with the bloodstain, and he eventually admitted to committing the crime.
Source: “How police found Gafoor,” BBC News, July 4, 2003, http://news.bbc.co.uk/1/hi/wales/3038138.stm (accessed April 15, 2010).
California’s Reversal on Familial Searching
The shift in the FBI policy also sparked disputes between states. Denver’s district attorney, Mitchell Morrissey, applied the interim FBI policy in familial searches to a cold case involving a Denver rape. The crime-scene DNA did not yield an exact match in CODIS, but a moderate-stringency search found what Morrissey believed could be a close family member who had been convicted of a felony in California. He requested information about the felon from California attorney general Jerry Brown in July 2007. On August 3, 2007, Brown reportedly denied the request, citing the need to protect the privacy of California felons who are not exact matches in CODIS. Brown noted that reporting on DNA near matches was beyond the scope of court opinions that authorize DNA database searches and could prompt a lawsuit.42
Five months later Rockne Harmon, a former senior deputy district attorney from Alameda County, California, and a strong proponent of familial searching, announced at the New York State Forensic Science Commission meeting on familial searching that California was about to change its policy on the release of partial-match information and unveil a full-scale familial searching program. He repeated these comments two months later at the March 2008 FBI symposium, and when he was asked by a representative from the Louisiana state lab why he was moving forward without legislative guidance, given the potential privacy and state constitutional concerns, he simply responded, “We’re doing it without legislation.”43
On April 14, 2008, the American Civil Liberties Union of Northern California filed a Public Records Act (PRA) request with the California Department of Justice (DOJ), asking for all records relating to the department’s policy regarding familial searching, as well as any plans to change its policy. Ten days later (the deadline for the California DOJ to issue a response) the agency released publicly a new “DNA Partial Match Policy.” A complete turnaround from Jerry Brown’s initial response against releasing partial-match information, the policy not only allows the release of information in the event of partial matches, in concert with the FBI’s interim policy, but it goes much further in explicitly allowing low-stringency searches. Requests for these searches are to be considered on a case-
by-case basis. The crime-scene profile must be single source (not a mixture), the search must produce a “manageable number of candidates,” and Y-STR analysis is required.44 A minimum of 15 shared alleles is required for partial-match follow-ups; no minimum threshold is given for low-stringency searches.
California is the first state to release an official policy directive on familial searching. Some states, such as Oregon, updated their CODIS operations manuals to incorporate the FBI’s interim policy on partial matches. Massachusetts and New York have regulations that explicitly address the issue of low-stringency searches, requiring that a minimum of four loci be provided for a forensic search against the DNA database. The intent of these regulations seems to be to address the issue of crime-scene sample degradation or limited sample availability rather than familial searches. However, Massachusetts provides that “the laboratory . . . may, at its discretion, request that a search be performed using fewer loci if there are scientific reasons which support using fewer than four loci in a particular case, including but not limited to the apparent presence of mixtures, sample degradation, limited sample availability, or the possible involvement of relatives.”45 Where other states have permitted familial searches, the threshold of similarity required for allowing follow-up is ambiguously defined and described in terms of such matches needing to “be very, very close” (Virginia) or is set at an arbitrary number of alleles (for example, 21 out of 26 in Florida).
Genetic Justice Page 10
