BOX 16.3 Gary Leiterman: Murderer or Victim of Cross-Contamination?
In March 1969 Jane Mixer, a 23-year-old University of Michigan law student, was murdered. No leads were generated for the case until 2002, when the Michigan State Police Crime Laboratory in Lansing processed the DNA evidence from the crime and found DNA from two men. A drop of blood from Mixer’s hand was found to match an individual named John Ruelas, who was in the database because he had been convicted of killing his mother. DNA taken from the pantyhose of the victim was found to match Gary Leiterman, who was in the database for having previously been convicted of fraud involving prescription drugs. Ruelas, it turned out, was only 4 years old at the time of the murder. Police could not find any link between the child Ruelas and Mixer, and no explanation was provided why his DNA was found on the victim’s hand 33 years after the crime was committed. At the same time, a review of the lab records revealed that DNA samples from both Ruelas and Leiterman were being processed for submission to CODIS in connection with other cases on the same day on which the old samples from the Mixer case were being analyzed. In addition, DNA from the victim was barely detectable on her pantyhose, indicating that significant degradation had occurred over the 33 years since the crime had been committed. However, Ruelas’s and Leiterman’s profiles did not show a similar level of degradation, indicating that they were unlikely to have been deposited at the same time as the victim’s DNA.
When these facts are taken together, it seems plausible that the detection of DNA from both Ruelas and Leiterman could have been due to contamination of the Mixer crime-scene evidence that occurred in the laboratory while the DNA samples were being processed. Certainly this seems the only likely explanation for the presence of Ruelas’s DNA. Nonetheless, Leiterman was convicted of Mixer’s murder in 2005.
Sources: Amalie Nash and Art Aisner, “DNA Evidence Key in 1969 Slaying Trial,” Muskegon Chronicle, July 22, 2005; William C. Thompson, “The Potential for Error in Forensic DNA Testing (and How That Complicates the Use of DNA Databases for Criminal Identification)” (paper produced for the Council for Responsible Genetics [CRG] and its national conference, “Forensic DNA Databanks and Race: Issues, Abuses and Action,” New York University, June 19–20, 2008); Theodore Kessis, “Report of Findings, People v. Gary Leiterman, No. 04-2017-FC,” http://www.garyisinnocent.org/web/CaseHistory/NewDNAFindings/tabid/58/Default.aspx (accessed April 28, 2010); People of the State of Michigan, Plaintiff-Appellee v. Gary Earl Leiterman, Defendant-Appellant, Court of Appeals of Michigan, no. 265821, decided July 24, 2007.
Myth of Objectivity
Two DNA Samples Either Match or They Do Not Match;
DNA Analysis Is Not Subject to Interpretation.
Even if DNA samples are collected, handled, and processed with the greatest care and errors are minimized to the highest extent possible, the results of DNA analysis are still subject to interpretation. The subjectivity of DNA analysis is largely unacknowledged in popular accounts.
In the early development of DNA forensic analysis, the output came in the form of an autoradiogram, which had some resemblance to a supermarket bar code. The bars, which represented the appearance of an allele at a specific locus, showed up in different intensities. When bar segments of the autoradiogram were very light, some interpreters of the data might neglect the bar, believing that it was an imperfection or an artifact. By neglecting the faded bar, the forensic DNA specialist could conclude that there was an exact or partial match; in the latter case he might report that the individual’s DNA profile was consistent with the profile of the crime-scene sample.
Advances in DNA testing methodology replaced autoradiograms with electropherograms that generally give a cleaner visualization of the alleles in a DNA profile. These graphs have peaks and numbers associated with each peak that identify the locus where the peak is found and give the number of STRs within an allele at the peak, and the height of the peak indicates the amount of DNA associated with the peak (see chapter 1). Although the new output from the DNA analyzers is a marked improvement over the output in autoradiograms, discretionary factors remain in the interpretation of the data.
First, there is no standard rule of thumb for how an analyst should report ambiguous results of DNA analyses. For example, where degradation has occurred, peak heights might be very low, and the profile might be considered incomplete. One analyst might decide that these measurements are spurious and unreliable and might report this result as “inconclusive,” while another might report a partial profile. Partial matches may not provide sufficient information to support the conclusion that it is extremely unlikely that someone other than the suspect would have the identical partial match, but they may provide sufficient information to exclude a person from a crime. If even a subset of alleles is found not to match that of a suspect, he or she cannot be the source of that sample. Unfortunately, there have been cases where lab analysts have failed to report analysis information that might have led to a different outcome in a case.
BOX 16.4 The Case of Robin Lovitt
In September 1999 Robin Lovitt was convicted and sentenced to death for the murder of a pool-hall manager in Arlington, Virginia. His conviction rested heavily on DNA evidence. A bloodstain was found on a jacket that Lovitt was wearing when he was arrested, several days after the crime was committed. The lab analyst reported that the results of the analysis of the stain were “inconclusive,” but the prosecutor argued that the stain came from the victim. In addition, bloodstains on the murder weapon matched the blood of the victim but also contained a single, additional allele that was shared by Lovitt. A DNA expert testified that this allele was shared by 19 percent of African Americans.
A closer look at the DNA evidence revealed that the lab analyst failed to report that the DNA analysis of the sample from Lovitt’s jacket produced a partial profile of five loci. All five of those loci matched Lovitt’s DNA. Therefore, there was no evidence that the blood on his jacket matched that of the victim, and if anything, the evidence that it matched Lovitt was exculpatory. In addition, the single extra allele that was found on the murder weapon was far more common in the population than was reported to the jury; according to the FBI’s population data, that allele is found among 33 percent of African Americans, 46 percent of Caucasians, and 40 percent of Hispanics.
Lovitt’s death sentence was reduced by Governor Mark Warner in 2005 to life imprisonment without the possibility of parole. The problems with the interpretation and presentation of the DNA evidence were never fully considered in any of the appeals.
Source: William C. Thompson and R. Dioso-Villa, “Turning a Blind Eye to Misleading Scientific Testimony: Failure of Procedural Safeguards in a Capital Case,” Albany Law Journal of Science and Technology 18 (2008): 151–204.
Cases that involve mixtures of DNA from two or more sources provide the most opportunity for ambiguity. In the face of a mixture, a forensic analyst usually attempts to separate the alleles so that the profiles of each of the contributors can be determined. Even for experienced forensic analysts, however, there may be several ways to sort the alleles among two or more contributors. If all the contributors are available for DNA testing, the sorting process can be carried out with a reasonable degree of accuracy. Some forensic scientists believe that the sorting of the alleles must be done before one has information about the profiles of possible suspects. To do otherwise could bias the interpretation.11 When the contributors to the DNA profile of interest are not all available or are in dispute, then certain assumptions can be made about the likelihood that the suspect’s profile is included among the evidence. The forensic investigator may choose one hypothesis that fits the suspect’s profile in the assemblage of alleles without disclosing to the jurors the other possible allele assortments that do not match the suspect. As Thompson and colleagues have noted:
By their very nature mixtures are difficult to interpret. The number of contributors is often unclear. Although the presence of three or more alleles at an
y locus signals the presence of more than one contributor, it often is difficult to tell whether the sample originated from two, three, or even more individuals because the various contributors may share many alleles.12
Misinterpretation of mixtures has resulted in false cold hits and even wrongful convictions. In analyzing mixed samples it is critical that the person engaged in the DNA analysis not have an interest or a stake in the outcome of the case and not be seeking to find a match among the alleles with a preexisting suspect’s DNA profile.
BOX 16.5 The Case of Josiah Sutton
In 2004 Josiah Sutton was exonerated after spending four and one-half years in prison for a rape he could not have committed. Sutton’s conviction rested almost entirely on the basis of DNA tests performed by the Houston Police Crime Laboratory. The lab claimed that a semen stain found in the back of the car where the rape occurred contained two profiles—Sutton’s and that of an unidentified man. In addition, the lab analyst testified that the DNA found in the sperm fraction of vaginal swabs and on the victim’s jeans “matched” Sutton’s. Reanalysis of the lab report showed that the semen sample came from a single source, and not from Sutton. In addition, the lab analyst exaggerated the significance associated with the inclusion of Sutton’s DNA profile in the mixed evidentiary sample by not reporting any statistics and repeatedly testifying about the uniqueness of each DNA pattern. It turned out that the chance of a coincidental match in this case was quite high: the frequency in the African American population of men who would be “included” in the vaginal sperm fraction was 1 in 15. Exposure of the errors in Sutton’s case led to a full-scale investigation of the Houston Crime Lab and the review of hundreds of cases involving DNA evidence.
Source: William C. Thompson, review of DNA evidence in State of Texas v. Josiah Sutton (District Court of Harris County, Case No. 800450), February 6, 2003.
Myth of Individuality
No Two People Can Have the Same DNA Profile. The Probability of a Coincidental Match Is Zero or Infinitesimally Small.
When the DNA profiles of two pieces of biological evidence “match,” it is often presumed that they must have come from the same source. But could the match have been coincidental? Do the police have an innocent person whose DNA profile happens to match perfectly the profile of the DNA left at the crime scene? What are the chances that two or more people share the same DNA profile and that the matching profiles do not represent DNA from the same individual?
There is a generally recognized assertion that, except possibly for monozygotic (identical) twins, no two people can have identical sets of 3 billion base pairs of DNA.13 In forensics, however, no person’s complete DNA is sequenced. In the United States 13 loci, as well as markers on the X and Y chromosomes for gender determination, are selected for DNA analysis. The question about coincidental matches reduces to this: what are the chances that more than one person (such as sibling pairs) will have the exact number of short tandem repeats (STRs) in the 26 alleles of the 13 loci used to profile their DNA?
There are three important principles used in developing the probability statistic that prosecutors use in court. The first principle states that the 13 loci are independent and thus are not linked in the population. This is based on a testable assumption that the loci are assorted randomly. The second principle, derived from the first, is that the probability that any individual has a particular array of STRs is given by the product of the frequency with which each allele appears in the population. The third principle states that individuals who are identified with similar racial, ethnic, or ancestral groups have a greater likelihood of allelic similarity in their DNA profiles, including the number of STRs at a locus in the chromosome, than individuals associated with other population groups. When a DNA profile match is found, the population frequencies of the alleles are most commonly determined from one of three population reference groups, Caucasian, African American, or Hispanic, on the basis of the perpetrator’s closest “racial” identity. The third principle allows forensic scientists to estimate the likelihood that two unrelated individuals have the same DNA profile for 13 loci or fewer.
Dan Krane describes a three-stage process that forensic laboratories use to determine the probability that the DNA taken from a random, unrelated individual in the population has the same profile as the evidence sample (the random-match probability, RMP).14 In step 1 the frequency of each allele in the DNA profile of interest is estimated in the reference database. As an example, at a particular locus, allele 1 (7 STRs) appears at a frequency of 3 percent, and allele 2 (12 STRs) occurs at a frequency of 4 percent. In the second step the frequency of each genotype is calculated by the formula 2 times p times q, where p and q are the frequencies of the two alleles in the genotype. The multiplier 2 comes from the fact that each allele can come from either the mother or the father. In the example here, the frequency of the genotype with alleles of frequencies .03 and .04, respectively, is 2 × .03 ×.04 = .0024. The frequency of the overall genotype of 13 loci is obtained by multiplying the frequencies of each locus.
To highlight the principles of probability underlying forensic DNA, consider the following example. Suppose that we have an urn filled with 1,000 balls, some red and some blue. Now imagine that we have selected 10 balls and found that 7 were red and 3 were blue. Can we assume that our next pick of 10 balls would give us the same number of red and blue ones? Obviously not. For one thing, we do not know whether we took a random sample of the balls in the urn. We also do not know whether the balls are distributed homogeneously, or whether all of the red balls are stacked at the bottom of the urn. But if we repeatedly selected 10 balls, we could calculate the average number of blue and red picks. If there were in fact 700 red balls and 300 blue ones, then our ratios in the picks of 10 would cluster around a mean of 7 red and 3 blue, although we would not get that ratio in every selection.
How do scientists know what the allele frequencies are in the population? How do they know how many people have a particular allele at locus 2? There is no direct way because neither the government nor scientists have the DNA profile of everyone in the world, and therefore they cannot calculate the exact frequency of the STR alleles of interest that exist in the entire population. Instead, scientists use convenience databases rather than a random sample of the population. The databases from which forensic scientists draw allelic frequencies could be a few hundred people whose DNA happened to be on hand when allele frequencies first needed to be determined. Because the databases are not a random sample of the population, in theory the probability estimate could either overestimate or underestimate the frequency of the alleles and thus give a false value for the chances of a coincidental match. As in the urn example, even without a random sample of the population, if we kept taking samples of allele frequencies from the population, we would eventually get an average that approaches the real frequency of the allele in the entire population.
But, unlike the example of balls in the urn, where we may know the distribution of red and blue balls, we do not know what the exact allele frequencies are in the population. Forensic scientists infer the actual allele frequencies from the small sample of people who do not represent a random sample of the population. If we drew the allele frequencies from a completely different population, it is likely that we would get a different set of frequencies. As in the case of the urn, if we chose enough population samples, we would expect to obtain a distribution of allele frequencies whose average would approach the average of the entire population.
Some forensic scientists argue that a random selection of the population for the purpose of obtaining allele frequencies is not necessary because even small reference groups will have allele frequency distributions at specific loci that are similar to those of the larger population. The analogy for the urn is that with sufficient mixing we can select 10 balls from the top (not random, but a convenience sample) and get 7 red and 3 blue balls.
The racial or ethnic background of someone who is the source of an evi
dence sample is often unknown or in dispute. As a result, allele frequencies from the three common racial groups named earlier are commonly used to attach a weight to any matches to such a profile. It is plausible to assume that the greatest chance of a coincidental match of a DNA profile would come from individuals who have similar phenotypes and therefore come from the same “racial” lineage. When an exact match is found between a DNA profile obtained from a biological specimen left at the crime scene and a DNA profile obtained from a suspect, the forensic investigator then determines the likelihood that some randomly chosen unrelated person whose DNA was not left at the crime scene would exhibit an identical DNA profile (the RMP).
Even with frequencies for each STR sequence in the range of 1 in 10 (1 person in 10 has the same number of repeats), the product rule rapidly yields a very low probability [(1 ⁄ 10)n, where n is the number of alleles]. The conventional wisdom within the forensic field is that the likelihood of a random match from 13 loci is inconceivable (perhaps one in a trillion) so long as the DNA is properly handled, no laboratory errors occur, and the individuals involved are not identical twins.
For example, if the frequency of each of the 26 alleles in a DNA sample as determined by the relevant population database is 1 out of 10, then the RMP that two nonidentical twins would have the identical profile is 2 × (1 ⁄ 10)26, or 1 in 50 septillion (1 septillion = 1024 or a trillion trillion). This point was made in Discover Magazine:
Genetic Justice Page 35
