Detection and Attribution of Organophosphate Pesticide Signatures

Organophosphate pesticides (OPPs) are a highly toxic class of chemicals still used for agricultural purposes in many countries. Their high toxicity and wide availability could make them attractive to terrorists or criminals for use as chemical threat agents (CTAs). 

The U.S. Department of Homeland Security came to Battelle for help in identifying new analytical and statistical methods for characterization of OPPs. Common analytical techniques such as gas chromatography coupled with nitrogen/phosphorus detection or mass spectral detection are able to accurately identify the parent materials in an unknown compound. However, they do not have the sensitivity needed for source attribution or forensic analysis. New methods were needed that would be able to detect trace impurities in order to distinguish characteristics such as the OPP source or manufacturing method and provide a distinct chemical “fingerprint” for use in forensic investigations.

Researchers at Battelle conducted a study to determine the chemical attribution signatures (CASs) for several commercially available OPPs. The compounds were analyzed using two-dimensional gas chromatography with time-of-flight mass spectrometric detection (GC×GC-TOFMS). Researchers then applied statistical pattern recognition techniques to the data collected in order to determine the unique chemical fingerprint of each sample. Replicate samples of chlorpyrifos, dichlorvos and dicrotophos were analyzed to identify CASs.

GC×GC-TOFMS provides much higher sensitivity than traditional analytical methods, giving it greater potential for forensic analysis. By using two dimensions of separation instead of one, it provides substantial increases in chromatographic separation, allowing for more detailed analysis. It also allows analysts to make tentative identification of unknown compounds in the absence of analytical standards. However, it can be difficult to sort through the large data sets produced by the analysis in order to find meaningful components or patterns. 

Because of the large number of predictor variables produced, data could not be analyzed using traditional statistical methods. Battelle researchers evaluated three statistical pattern recognition methods—Random Forest, Lasso and Elastic Net—to determine which method was most effective in classifying the compounds.