The jpoc five minute guide to DNA fingerprints
What are DNA fingerprints? How are they used? What are the issues and problems concerning them?
Everyone has heard of DNA fingerprinting, the revolutionary technique that has give forensic science a powerful tool in the identification of criminals. But what is really behind it?
First of all, consider what is your DNA. DNA is a piece of molecular engineering made by joining large numbers of chemical building blocks together. Imagine a tower made of Lego bricks of three different colours. The sequence of those colours carries information just as does the sequence of building blocks in a length of DNA. With a few exceptions, every cell in your body contains a copy of your DNA. These copies are all almost identical. Changes occur as a result of copying errors or damage from things like cigarette smoke or radiation. Think of these changes as being like the effects of a fifth generation photocopy. It will not look exactly the same as the original but there is no chance that you would mistake a fifth generation photocopy of the front page of today's Times for the front page of yesterday's edition.
Unless you have an identical twin, the information in your DNA is unique to you. It is vanishingly unlikely that there has ever been another human being with exactly the same DNA as you.
It is very easy to leave traces of your DNA behind you. In semen or blood, in dead skin cells scratched from your face in a fight, even in dead cells in saliva when you lick a stamp. (The truely paranoid already make a point of not licking stamps and envelopes.) Sometimes, those traces can be analysed to identify the person who left them behind.
Now, if your DNA is unique, you might think that it would be a pretty fool proof way to identify a person but you would be wrong. To explain why, I will make an analogy to car registrations.
If you witness a car accident and one driver just drives off at high speed, you might try to record the registration number of the vehicle. If you have that and it is correct, then you have a unique identification of the offending vehicle and you just need to look it up on a computer. But imagine that, instead of a few letters and numbers, a car registration was made up of one thousand letters and that there was no national computer system that stored all of the car registration numbers. What would you do then?
Perhaps you would have noted that the first three digits in the registration were 528. Now, the authorities could try to search for a car with such a registration but, when they found such a car, the probability that it was the one in the accident would vary depending on how it was found.
For example, suppose that, after the accident, you saw the car drive off and turn into a car park. The police search the car park and find one car whose registration begins with 528. As long as you had seen no car leave the car park before they arrived, they can be completely certain that this car was the one in the accident.
Now, suppose that the police find two such cars in the car park, now your information is almost useless. Unless there is other evidence, they can only say that it is fifty fifty as to which of the two cars was involved in the accident.
Suppose that you did not see that car enter a car park and the police can do no more than to round up all of the cars in the area and check their numbers. Perhaps a search of the area turns up one car with the correct first three digits of the registration number. Does that make it the car in the accident? Well, what if the area contained one thousand cars. You might expect to find one car whose number began 528 just by chance so, in this case the evidence that you have given to the police is of no use at all in identifying the car. After all, the police could have searched a slightly wider area and they might have found ten thousand cars and a dozen with the correct sequence of digits. They have no grounds to say that they have identified the car.
The situation with DNA fingerprinting is very much like this. It is just not possible to identify the complete sequence of codes in the DNA of a sample of, say Semen so approximate methods are used. To return to the idea with the car registration, suppose that you actually had the whole number, but you could not look at all of the individual digits. Suppose though that you knew that every registration number contained the digits 36653481 at a position about one third of the way from the start but that the exact position varied from one car to another. Now, if you are able to measure where this sequence occurs in the number of the car in the accident, you can compare that position with the position of the same sequence in all of the cars that you want to test. This is just what is done in DNA fingerprinting.
Some sequences of DNA are common to us all. Without them, we would not be human. It is possible to make a chemical that will first stick exactly to a certain sequence in somebody's DNA and secondly cut the DNA into two pieces at that point. By choosing a large number of such sequences, cutting a person's DNA at these points and then measuring the resultant lengths of the fragments, you have a good way of identifying someone if everything else is in your favour.
Now, the DNAfingerprinting test kits that have been developed have different selections of cutting points and each gives its own "accuracy" which is best described as indicating say "one chance in half a million that the test matched by chance." Now one chance in half a million is certainly "beyond reasonable doubt" for convicting a person of a crime but, the accuracy of the test does not necessarily reflect the guilt of an accused person. Think back to the situation with the car numbers and then apply the same logic to the case with, say, a rape.
Consider first, the case where there are only a very limited number of men who had the opportunity to commit the offence. If there were twenty men in an hotel when a woman guest was raped and murdered then, if DNA fingerprinting throws up a match, that is pretty strong evidence. If the test itself has one chance in 500.000 of showing a match from two different people then, if one man in the group of twenty matches, the odds against this group of twenty having a match just by chance are 25,000 to 1 which is a pretty convincing case for guilt. However, suppose that the case was not so easily solved and the crime happened in a large city. There have been cases recently where police have launched mass screening programmes and now, things are very different. If it is the case that a test has one chance in half a million of throwing up a random match and if the police test every man in a medium sized town, then the odds that a match is random may come down to 5 to 1 against. Now that is hardly beyond reasonable doubt.
There is also a considerable racial bias in these tests. So, our sample test that has one chance in half a million of a random match against the general population may look very different if the sample is restricted to a narrowly defined group. Suppose that a victim of an attack in a large city identifies her attacker as West Indian and the police only test people from that group. A sample of as few as 10,000 men might throw up a match at random. That is hardly conclusive evidence.
So, to summarise, DNA fingerprinting is very useful in certain circumstances but the limitations of the techniques mean that it is not an accurate weapon for mass screening programmes as is sometimes claimed.