High-speed DNA fingerprinting
[March 5, 2007]
Criminals beware. A team of chemists from Cuba and Brazil has developed a novel automated method for producing DNA fingerprint profiles. This automated method is almost as accurate as a human DNA fingerprinting expert, but is a great deal faster.
DNA fingerprinting is a technique for identifying specific individuals from unique segments of their genome and is now so sensitive that it can work with tiny amounts of DNA extracted from bodily materials such as hair and saliva. As such, it is regularly used by police forces around the world, both in their current investigations and to provide new evidence in long unsolved or contentious cases.
Modern DNA fingerprinting techniques work by studying small segments of DNA known as short tandem repeats (STRs), which are non-coding DNA sequences (part of the so-called 'junk' DNA). As their name suggests, STRs consist of repeated sequences of between two and six DNA bases. These STRs appear at set points (known as loci) on the human genome, but the precise number of repeated sequences at each loci varies between different people and ranges from four to around 40.
While two people may have the same number of repeats at a single loci, the chances of them having the same number of repeats at numerous loci is astronomical. This pattern of repeated sequences can therefore act as a unique barcode (or fingerprint) for an individual.
In practice, DNA fingerprinting involves extracting the DNA from a bodily sample and amplifying the STRs at 13 loci using the polymerase chain reaction (PCR). These amplified STRs are then separated by electrophoresis, with small STRs travelling faster than larger STRs, allowing the length of the STR at each loci to be ascertained.
Traditionally, gel electrophoresis has been used to separate the STRs, with each of the 13 STRs forming individual spots at different points across the gel. But this process is hugely time-consuming and requires a DNA fingerprinting expert to generate the profile, which involves detecting the 13 STR spots and assessing their length. It is therefore being superseded by capillary electrophoresis, which although more expensive is much faster and more automated. But now a team led by Isneri Talavera Bustamante from Cuba's Advanced Technology Application Center may have struck a blow for gel electrophoresis, by developing an automated method for detecting and assessing STR spots.
To do this, Bustamante and her team realised that they first had to discover the best way to characterise the spots for subsequent computer processing. Particularly as the gels used for DNA fingerprinting usually contain a variety of spots, only some of which represent STRs. So they came up with a list of 14 numerical descriptors, which included values for area, compactness and height-width ratio, and then applied these descriptors to sample gels containing a total of 638 spots.
They analysed the results from this study using both principal components analysis, which identifies those factors most responsible for the variation in a sample, and a decision tree, which can measure how well a specific attribute is able distinguish between different examples. Their aim was to find which of these 14 descriptors was best able to distinguish STR spots from non-STR spots.
Combining the findings from both analytical techniques, they identified six descriptors that proved most effective at distinguishing the STR spots. They then used these descriptors in conjunction with a classification algorithm known as support vector machine to develop a quick and easy method for identifying each of the 13 SPR dots on a gel. Finally, they developed a fairly simple process for automatically calculating the lengths of these STRs.
Comparing their automated system with a DNA fingerprinting expert, Bustamante and her team found that it was only slightly less accurate at generating profiles, achieving a success rate of 97%. But their automated system produced the profiles in a fraction of the time, taking 15 minutes rather than 20 days.
AI
