When extractions go toxic
By extracting information from a freely-available chemical database, Italian forensic scientists have come up with a simple but highly effective method for identifying unknown toxicological compounds in biological samples.
A central component of forensic analysis and drug testing, toxicological analyses of biological samples have traditionally been conducted using gas chromatography/mass spectrometry (GC/MS), with compounds identified by comparing the resultant mass spectra with reference mass spectra in spectral databases.
Although effective at identifying toxicological compounds, this approach has certain limitations, such as the fact that GC/MS is not very good at detecting polar and non-volatile compounds. Liquid chromatography (LC) coupled with MS offers a more flexible alternative, able to identify both polar and non-volatile compounds. However, it suffers from an inability to produce as detailed mass spectra as GC/MS, which has prevented the construction of large spectral databases for LC/MS. According to Aldo Polettini of the University of Verona, LC/MS databases generally contain spectra for only around 1200 compounds, compared to spectra for tens of thousands of compounds in GC/MS databases.
In recent years, however, a type of MS known as time-of-flight (TOF) has opened up another way of identifying toxicological compounds with LC/MS. TOF-MS measures the masses of charged molecules based on the time they take to travel along a chamber to a detector under the influence of an electric field, with smaller molecules travelling faster than larger molecules. This allows it to make measurements of molecular mass that are so accurate that they can be used to determine a compound’s chemical formula, and thereby also its identity.
The problem is that there aren’t any major databases containing information on molecular mass and chemical formulae specifically for toxicological compounds, so Polettini decided to create one. The easiest way to do this is to extract data for toxicological compounds from an existing chemical database and Polettini chose to do this with the US National Institutes of Health’s PubChem Compound. This is freely-available on the internet and comprises around 10 million entries, each of which contains information on a compound’s molecular mass and chemical formula.
Together with colleagues, Polettini created a subset of these entries by extracting all the compounds that could be classified as toxicological and then screening them based on their molecular mass and whether they contain elements such as hydrogen, nitrogen or fluorine. This resulted in a subset database containing entries for 50,500 toxicological compounds, including many drug molecules, both pharmaceutical and recreational, pesticides and poisons, as well as metabolites.
‘It contains a large number of metabolites,’ explains Polettini, ‘including glucuronides, which are very important in general unknown screening, especially when metabolite-rich biological matrices are used (e.g. urine).’
Once the database was up and running, all Polettini and his team needed to do to determine the chemical formula of an unknown toxicological compound was to match the molecular mass revealed by TOF-MS with the matching mass in their database. Testing this approach on hair, blood and urine samples from subjects that had taken pharmaceutical or recreational drugs, they found that they were able to identify a whole host of relevant toxicological compounds.
Predictably, the only problem they found was that a specific molecular mass can match more than one chemical formula and a specific chemical formula can match more than one toxicological compound. But the correct compound could usually be pinpointed by simply taking other available information into account, such as some of the spectral data produced by TOF-MS.
A molecule’s retention time in LC can also offer a way to chose between competing identities. To this end, Polettini is now attempting to enhance the database by incorporating an algorithm for estimating the retention time for proposed compounds. These estimates can then be compared with the actual retention time of the detected compound to help reveal the correct identity.
