DATABASES

• Jurisdiction: Australia

DATABASES

Overview ....................................................................................................... [80A.1500]

Sample selection ........................................................................................... [80A.1520]

Making estimates from population samples ................................................. [80A.1540]

Sampling uncertainty..................................................................................... [80A.1560]

[80A.1500] Overview

To draw inferences about members of a population, a sample of the population must be selected. Most forensic science laboratories maintain databases of DNA profiles obtained from a sample of the population, and use these databases for the purpose of estimating random match probabilities.

[80A.1520] Sample selection

Ideally, the sample should be selected randomly from the population in question. This means that the relevant population must be identified and that all members of the population must have an equal chance of being selected. In practice, this cannot be done, because there is usually very little information about the population from which the actual offender came, and even if the population could be identified, there would have to be some means of randomly selecting individuals from it and compelling them to provide DNA profiles.

Most laboratories collect DNA profiles from a variety of sources, including people involved in criminal investigations (victims, husbands or boyfriends, suspects, convicted offenders), their own staff, donors to blood banks or people involved in disputed paternity testing. Although these are clearly not random selections, there are several arguments that have been put forward to support the idea that such databases are representative of the populations from which they are derived.

The first argument is that the samples are unlikely to be biased. When samples collected from people in different places or different groups within the same population are compared, it is generally found that there are no statistically significant differences between different samples: Weir et al (2004); Balazs et al (1992). If the samples in these studies were biased, they must all be biased in the same way. A simpler, and more probable explanation, is that the databases are all representative samples of the larger population.

The second argument is that no genetic characteristic has been identified that might predispose a person having that characteristic to be involved in criminal investigations. Modern DNA profiling techniques specifically involve the identification of genetic variants that are located in non-coding regions of the genome (either between genes, including loci whose designations begin with D (eg, D1S80; Nakamura, Carlson, Krapcho and White, 1988; Kloosterman, Budowle and Daselaar, 1993), or in introns within genes, such as vWA and TH01: Kimpton et al (1993). If these regions have some function, that function is unknown: Mattick (1994; 2001). Therefore, it is claimed, possession of particular variants is unlikely directly to confer on a person any genetic or phenotypic characteristic that would make them more likely to be included in a database. If there is any bias, it could only arise as a result of indirect mechanisms. For example, a database may contain large numbers of blood donors, who may be more likely to be of a particular religion, which might be overrepresented by a particular subgroup. It seems likely that the genetic make-up of the people in the databases is not vastly different from the genetic make-up of people in the general population.

Further support for the proposition that DNA databases are representative of the populations from which they are drawn comes from the success of attempts to explore the relationship between populations, leading to the construction of evolutionary trees from DNA databases: see, for example, Weir (2004); Walsh et al (2007); Buckleton, Bright and Taylor (2016). If the databases were vastly unrepresentative, this should not happen.

Two other issues arise when compiling databases: duplicates and relatives. The inclusion of duplicate profiles from the same person will result in spurious indications of associations between alleles and inbreeding in the population: Buckleton et al (2001). To avoid including the same person twice, it is essential that some identifying information is retained. It is not necessary to keep the identifying information with the profile, as long as it is possible to determine whether a...