|
Arsonists may be captured just that little bit quicker, thanks to research conducted by chemists from the University of Central Florida's National Center for Forensic Science (NCFS). These chemists have shown that applying a mathematical technique known as covariance mapping to gas chromatography-mass spectrometry (GC-MS) data provides a fairly simple way to tell whether or not two petrol (gasoline) samples are the same. One of the first questions that forensic scientists need to answer when called to investigate a case of arson is what kind of flammable substance was used to start the fire. But determining the identity of a flammable substance, which can mean a specific brand of petrol, from fire debris is far from easy, even using sensitive analytical techniques such as GC-MS. This is because factors such as weathering (evaporation), biological degradation and small variations in separation techniques can make it difficult to spot the signal of a specific flammable substance in the complex spectra produced by the debris. Sometimes, forensic scientists are asked the seemingly easier question of whether two samples of fire debris contain the same flammable substance. At the moment, however, answering this question still generally involves the difficult task of identifying the flammable substance. A much simpler method would be merely to compare the spectra produced by the two samples to see how much they differ: spectra that differ beyond a certain threshold would then imply different flammable substances. Now, a team of chemists from NCFS, led by Michael Sigman, have demonstrated that covariance mapping can form the basis for just such a method. Covariance is a measure of the degree to which two random variables vary together; covariance mapping is a way to display the covariance of the many different variables in complex samples in the form of a two dimensional matrix. To apply covariance mapping to GC-MS data, Sigman and his team first develop a matrix of individual ion abundances for each sample by plotting the time of each mass scan against the detected mass-to-charge ratios. To create a covariance matrix, they then simply multiply this sample matrix by its transpose (in which rows in the original matrix become columns and columns become rows). To compare the difference between the covariance matrices of two different samples, Sigman and his team then work out the distance between the matrices. This is done by calculating the difference in magnitude between each of the comparable matrix values (which have been normalised such that they add up to one in each matrix), summing the differences together and then dividing by two. This produces a single value between zero and one, with zero representing identical samples and one representing completely different samples. But in the real world even supposedly identical samples will have distances slightly greater than zero, because of slight differences in their composition and small variations in the analyses. So the trick is to analyse identical samples to discover the distance value generated by this unavoidable variation. Any distance between two samples that is significantly above this value then indicates that the samples are different. Using this method, Sigman and his colleagues first showed that it could distinguish between unweathered or lightly weathered petrol samples and heavily weathered samples, as well as between different classifications of petrol distillates, such as light, medium and heavy. More recently, they have shown that covariance mapping can also distinguish between 10 different brands of petrol from oil companies such as Mobil, BP and Shell. Sigman and his colleagues are now investigating other potential forensic uses of covariance mapping. 'We are exploring the use of covariance mapping in several areas of forensic analysis that involve comparisons of samples from questioned and known origins,' he told separationsNOW. Related links:
The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd. |
 |
|
|
|
|
