Silent Witness - Volume 8, Number 1, 2003

Meeting Defense Challenges to DNA Evidence

by Steve Hogan & Steve Swinton, New York State Police

Some prosecutors have experienced that feeling of dread, panic and powerlessness when their expert witness is not faring well during cross-examination. These moments tend to happen when the expert witness testimony involves molecular biology or population genetics. What is most frustrating in forensic DNA cases is that the prosecutor knows that good science is being portrayed incorrectly as bad science.

Never before have prosecutors had a more powerful tool at their disposal for determining the identity of persons who commit crime. The use of PCR-based testing of Short Tandem Repeat (“STR”) locations on the DNA molecule has revolutionized forensic science. Additionally, the STR DNA tests now used in crime laboratories are highly “discriminating.” The resulting profiles often allow for random match probability estimates that are less than 1 in a quadrillion.

There are few, if any, techniques in the history of forensic science that have been more thoroughly scrutinized, validated, and tested than forensic DNA testing. The underlying science of DNA testing is virtually unassailable. For this reason, few defense experts will dare to challenge the validity of DNA amplification or the detection of genetic variation using electrophoresis.

Prosecutors can expect defense challenges to the population statistics associated with a DNA match. Some of these attacks will focus on issues such as the use of the “product rule” to calculate profile frequency statistics, and the database size and/or the populations that were either included, or not included, in the database. A final issue is misstating the meaning of the statistics. Those defense attempts to undermine the significance of the DNA match and suggested responses are outlined in this article.

Forensic Statistics

Statistics are used to interpret the significance of a “match” between crime scene evidence and a reference sample (such as a suspect control sample). The random match probability statements that appear in many laboratory reports are statistical estimates of the expected frequency of a particular profile in the population. For example, if there were a “match” between the blood type (e.g., type O positive) of a crime scene sample, and the O positive blood type of a suspect, this match would not be particularly probative, since a large percentage of the population has type O positive blood. By contrast, if there were a match at each locus for every one of the two possible alleles at that locus, logic suggests that such a match would be extremely probative. Statistics demonstrate the probability of finding a coincidental match in the population.

The “Product Rule Should Not Be Used” Argument

Frequently, the battle between the prosecution and the defense when statistics are challenged boils down to the following: the prosecution wants to use the “product rule,” while the defense expert will likely argue in favor of “the counting method” or the “ceiling principle.”

Defense: “You’ve taken these numbers and multiplied them together for no good reason and the result is prejudicial to my client.”

Response: There’s a very good reason why the forensic scientist “multiplies the numbers” (actually the genotype frequencies at each locus). The product rule is a mathematical rule: the frequency of occurrence of several independent events is equal to the product of their individual frequencies. For example, if you flip a coin, the probability of getting “heads” is one half. The probability of getting “heads” three times in a row is one half multiplied by one half multiplied by one half, or one in eight. Since a person’s profile at one DNA locus has no bearing on the person’s profile at another locus, the frequency of occurrence of the entire profile equals the product obtained by multiplying the frequencies at each individual locus.

In 1996, the National Research Council (NRC), a division of the National Academy of Sciences, endorsed the product rule to calculate DNA random match probability. The NRC study, The Evaluation of Forensic DNA Evidence (also known as “NRC II”), is considered by most scientists in the field to be the definitive pronouncement on the issue of DNA match statistics. In NRC II, the experts considered and specifically rejected many of the alternative methods of calculating DNA random match probabilities.

Defense: “But your ‘product rule’ doesn’t take into account ‘population substructure,’ so the resulting frequency is too low.”

Response: Our application of the product rule does account for population substructure. In fact, it follows the very conservative approach recommended by the National Research Council in NRC II.

Note: Population substructure is another way of referring to inbreeding. All populations, no matter how defined, are inbred to some degree. The “theta factor” approach accounts for this in a way that favors the defendant by replicating the higher frequencies that are expected when there is inbreeding. Make sure the statistics applied in your case were calculated using one of the two approaches advocated by the NRC II.

The “Your databases are too small” argument

Defense: “You took a database of 200 people and used it to calculate a probability of one in a trillion. Your databases are too small.”

Response: The databases are not too small. They have been analyzed statistically, taking into consideration the number of loci tested and the number of alleles at each locus, and determined to accurately reflect the frequencies of alleles in the population as a whole. A database must be a sufficient size to calculate meaningful statistics; however, the size required is only a few hundred samples. Public opinion polls, for example, do not question everyone. Instead, they carefully sample a population to obtain an accurate estimate of the opinion of that population. [Fingerprints provide another example. Human fingerprints have been scientifically determined to be “unique” even though only a portion of the population of the world has been fingerprinted.] When sampling is done appropriately, the results are highly accurate.

Note: Make sure the databases used have been peer reviewed or subjected to extensive statistical analysis to ensure reliability.

Mixtures can’t be interpreted accurately.

Defense: “Maybe you can draw a conclusion when you’re dealing with a simple, single-source profile, but mixtures are too complicated to interpret, and reliable conclusions cannot be made from them.”

Response: Mixtures can be more complicated to interpret than single-source profiles, but the vast majority can be reliably interpreted and statistically evaluated. In fact, the peer-reviewed, supporting literature on this topic is quite extensive.

Note: It is possible that a particular mixed DNA sample will not be amenable to meaningful interpretation. This situation can arise if there are a large number of contributors (generally four or more) or if the sample is badly degraded. Mixtures of four or more individuals are uncommon. Badly degraded samples are more common.

It should be noted that after a forensic scientist issues a laboratory report indicating that a mixture is present and identifying the number of possible contributors to the mixture, the prosecutor may be asked to obtain control samples from possible contributors to that mixture.

The “My Client is a Green Zombie” Argument

Defense: “These statistics you’ve calculated are all very nice, but since my client is not a member of any of the ethnic groups you used (or is of mixed ethnicity), they are inapplicable and prejudicial to my client.” In other words, because the population database does not include a specific ethnic group, the random match probability is inaccurate.

Response: In making the random match probability calculations, the forensic scientist proceeds from the assumption that the defendant is not the source of the DNA at the crime scene. (The only assumption the forensic scientist makes is that the source is a human.) When a large number of loci are tested and very small probabilities estimated, the difference race or ethnic groups may produce is insignificant: any probability beyond six billion, is larger than the world’s population or any subgroup population. With an abundance of caution, several majority populations are used in order to provide varied random match probabilities: Caucasian-American, African American, Southwestern U.S. Hispanics and Southeastern U.S. Hispanics. By including 4 populations in a large enough database and testing 13 loci, the random match probabilities generated account for subtle variations that could occur in all population groups. Moreover, unless there is credible evidence to suggest that the actual contributor of the DNA is not the defendant but, like the defendant, is a “Green Zombie” (a member of the same un-represented ethnic group as the defendant), then the exclusion of the absent ethnic population (Green Zombies) is irrelevant and the challenge groundless.

Note: If other evidence in the case indicates that the contributor of the crime scene DNA must have been a Blue Plutonian (or whatever actual group to which the defendant allegedly belongs), this fact should be brought to the attention of the DNA analyst. There are an enormous number of databases available, at least one of which will likely represent the group to which the defendant belongs. Statistics can be calculated for this group before trial.

The Defense Fallacy

Defense: “That number you mentioned, one in a million. That means that in a city of 10 million people, there are nine people, other than the defendant, who are equally likely to have contributed the DNA found at the crime scene.”

Response: While [in this example] there may be nine people, other than the defendant, whose DNA profile matches that of the crime scene DNA, it is incorrect to say that any of those nine people is “equally likely” to have contributed the DNA at the crime scene. Other evidence with the DNA led to the defendant being charged. The DNA match is simply only one piece of all the evidence pointing to the defendant’s guilt.

Note: The random match probability will often be much lower than one in a million. If, for example, the random match probability is one in a trillion (not at all uncommon), only a city of two trillion (about 300 times the population of the Earth) would be expected to include another person with a profile matching the crime scene evidence.

The number of convicted offender profiles in the national DNA database (the Combined DNA Index System or CODIS) is currently around one million. A CODIS database hit means that when more than one million profiles were searched, only the defendant’s profile matched the profile obtained from the crime scene evidence. Moreover, the CODIS system is designed to generate leads. Cases are not normally brought unless there is other evidence connecting the person identified in the hit to the victim and the crime scene.

Conclusion

Prosecutors today often have the benefit of outstanding DNA testing technology. These suggestions may help frontline prosecutors fend off attacks on DNA match evidence and help fact finders reach just verdicts.

Click here for the full length article with footnotes.