DNA profiling can be used for identifying individuals and determining relationships

DNA profiling can be used for identifying individuals and determining relationships

17.4. DNA profiling can be used for identifying individuals and determining relationships

We use the term DNA profiling to refer to the general use of DNA tests to establish identity or relationships. DNA fingerprinting is reserved for the technique invented by Jeffreys et al. (1985) using multilocus probes. For more detail on this whole area, the reader should consult the book by Evett and Weir (Further reading).
17.4.1. A variety of different DNA polymorphisms have been used for profiling

DNA fingerprinting using minisatellite probes

These probes contain the common core sequence of a hypervariable dispersed repetitive sequence GGGCAGGAXG, first discovered by Jeffreys et al. (1985) in the myoglobin gene (see Section 7.4.2). When hybridized to Southern blots they give an individual-specific fingerprint of bands (Figure 17.19). Their chief disadvantage is that it is not possible to tell which pairs of bands in a fingerprint represent alleles. Thus, when matching DNA fingerprints, one matches each band individually by position and intensity. Other hypervariable repeated sequences have been used in the same way, for example those detected by the synthetic oligonucleotide (CAC)5 (Krawczak and Schmidtke, 1998). top link
DNA profiling using single-locus minisatellite markers

Minisatellite probes recognize single-locus variable tandem repeats on Southern blots. Each probe should reveal two bands in any person's DNA, representing the two alleles. Profiling is based on four to ten different polymorphisms. These probes allow exact calculations of probabilities (of paternity, of the suspect not being the rapist, etc.), if the gene frequency of each allele in the population is known. For matching alleles between different gel tracks, the continuously variable distance along the gel has to be divided into a number of ‘bins'. Bands falling within the same bin are deemed to match. It is imperative that the criteria used for judging matches in each profiling test should be the same binning criteria that were used to calculate the population frequencies of each allele. The binning criteria can be arbitrary within certain limits, but they must be consistent. Minor variations within repeated units of some minisatellites potentially allow an almost infinite variety of alleles to be discriminated, so that the genotype at a single locus might suffice to identify an individual (Jeffreys et al., 1991).top link
DNA profiling using microsatellite markers

Microsatellite polymorphisms (Section 7.4.3) are based on short tandem repeats, usually di-, tri- or tetranucleotides. They have the advantages over minisatellites that they can be typed by PCR and that discrete alleles can be defined unambiguously by the precise repeat number. This avoids the binning problem and makes it easier to relate the results to population gene frequencies.top link
The use of Y-chromosome and mitochondrial polymorphisms

For tracing relationships to dead persons, Y-chromosome and mitochondrial DNA polymorphisms are especially useful because of their sex-specific pattern of transmission. An interesting example was the identification of the remains of the Russian Tsar and his family, killed by the Bolsheviks in 1917, by comparing DNA profiles of excavated remains with living distant relatives (Gill et al., 1994).top link
17.4.2. DNA profiling can be used to determine the zygosity of twins

In studying nonmendelian characters (Chapter 19), and sometimes in genetic counseling, it is important to know whether a pair of twins are monozygotic (MZ, identical) or dizygotic (DZ, fraternal). Traditional methods depended on an assessment of phenotypic resemblance or on the condition of the membranes at birth (twins contained within a single chorion are always MZ, though the converse is not true). Errors in zygosity determination systematically inflate heritability estimates for nonmendelian characters, because very similar DZ twins are wrongly counted as MZ, while very different MZ twins are wrongly scored as DZ.

Genetic markers provide a much more reliable test of zygosity. The extensive literature on using blood groups for this purpose is summarized by Race and Sanger (1975). DNA profiling is nowadays the method of choice. The Jeffreys fingerprinting probe allows a very simple test - samples from MZ twins look like the same sample loaded twice, and samples from DZ twins show some differences. An error rate could be calculated from empirical data on band sharing by unrelated people, using some defined binning strategy (see above).

When single-locus markers are used, if twins give the same types, then for each locus, the probability that DZ twins would type alike is calculated. If the parents have been typed, this follows from mendelian principles; otherwise the probability of DZ twins typing the same must be calculated for each possible parental mating and weighted by the probability of that mating calculated from population gene frequencies. The resultant probabilities for each (unlinked) locus are multiplied, to give an overall likelihood PI that DZ twins would give the same results with all the markers used. The probability that the twins are MZ is then:

where m is the proportion of twins in the population who are MZ (about 0.4 for like-sex pairs). Sample calculations are given in Appendix 4 of Vogel and Motulsky (Further reading).top link
17.4.3. DNA profiling can be used to disprove or establish paternity

Excluding paternity is fairly simple - if the child has a marker allele not present in either the mother or alleged father then, barring new mutations, the alleged father is not the biological father. Proving paternity is, in principle, impossible - one can never prove that there is not another man in the world who could have given the child that particular set of marker alleles. All one can do is establish a probability of nonpaternity that is low enough to satisfy the courts and, if possible, the putative father.

DNA fingerprinting probes have been widely used for this purpose (Figure 17.19). Bands must be binned according to an arbitrary but consistent scheme, as explained above, to decide whether or not each nonmaternal band in the child fits a band in the alleged father. Then if, say, 10/10 bands fit, the odds that the suspect, rather than a random man from the population, is the father are 1:p10, where p is the chance that a random man from that population would have a band matching a given band in the child. Even for p = 0.2, p10 is only 10-7. Single-locus probes allow a more explicit calculation of the odds (Figure 17.20). A series of four to ten unlinked single-locus markers can give overwhelming odds favoring paternity if all the bands fit.top link
17.4.4. DNA profiling is a powerful tool for forensic investigations

DNA profiling for forensic purposes follows the same principles as paternity testing. Scene-of-crime material (bloodstains, hairs or a vaginal swab from a rape victim) are typed and matched to a DNA sample from the suspect. If the bands don't match, the suspect is excluded. One of the most powerful applications of DNA profiling is for preventing miscarriages of justice. If the bands do all match, the odds that the criminal is the suspect rather than a random member of the population can be calculated, based on the allele frequencies in the population. Of course, if the alternative were the suspect's brother, the odds would look very different. The fate of DNA evidence in courts provides a fascinating insight into the difference between scientific and legal cultures. There are at least three stumbling blocks for DNA data.

* The jury may simply not believe, or perhaps choose to ignore, the DNA data, as evidently happened in the OJ Simpson trial. A fascinating account of the DNA evidence is given by Weir (1995).
* The jury may be led into a false probability argument, the so-called Prosecutor's Fallacy. Suppose a suspect's DNA profile matches the scene-of-crime sample. The Prosecutor's Fallacy confuses two different probabilities:
1. the probability the suspect is innocent, given the match;
2. the probability of a match, given that the suspect is innocent.

The jury should consider the first probability, not the second.

Using Bayesian notation (Box 17.2), with M = match, G = suspect is guilty, I = suspect is innocent, we want to calculate PI | M, and not PM | I. If the suspect were guilty, the samples would necessarily match: PM | G = 1. Population genetic arguments might say there is a 1 in 106 chance that a randomly selected person would have the same profile as the crime sample: PM | I = 10-6. Suppose the guilty person could have been any one of 107 men in the local population. If there is no other evidence to implicate him, he is simply a random member of the population and the prior probability that he is guilty (before considering the DNA evidence) is PG = 10-7. The prior probability that he is innocent is PI = 1–10-7, ~ 1. Baye's theorem tells us that

The prosecutor would no doubt be happy to see the jury use 106 instead of 0.9 for the probability that the suspect is innocent! Given the Bayesian argument, it is clear that a forensic test needs PM | I to be 10-10 or less if it is to be able to convict a suspect on DNA evidence alone.
* Objections may be raised to some of the principles by which DNA-based probabilities are calculated.
1. The multiplicative principle, that the overall probabilities can be obtained by multiplying the individual probability for each band or locus, depends on the assumption that bands are independent. If the population were actually stratified into reproductively isolated groups, each of whom tended to have a particular subset of bands or alleles, the calculation would be misleading. This is serious because it is the multiplicative principle that allows such exceedingly definite likelihoods to be given.
2. For single-locus markers, the probability depends on the gene frequencies. DNA profiling laboratories maintain databases of gene frequencies - but were these determined in an appropriate ethnic group for the case being considered? Taken to extremes, this argument implies that the DNA evidence might identify the criminal as belonging to a particular ethnic group, but would not show which member of the group it was who committed the crime.

These issues have been debated at great length, especially in the American courts. Both objections are valid in principle, but the question is whether they make enough difference to matter. General opinion is that they do not. It would be ironic if courts, seeing opposing expert witnesses giving odds of correct identification differing a million-fold (105:1 versus 1011:1), were to decide that DNA evidence is hopelessly unreliable, and turn instead to eye-witness identification (odds of correct identification < 50:50).top link