Discrimination of drug isomer EI mass spectra using consensus peak likelihood estimates

Mass spectrometry is widely used in the identification of chemical compounds, particularly for seized drug analysis by forensic crime labs. Structurally similar compounds such as isomers generate similar mass spectra. Measurement variability causes such spectra to be indistinguishable under traditional similarity scoring systems. Previous work has shown that the use of replicate mass spectra can overcome this problem, with the help of a novel clustering and scoring algorithm called the high dimensional consensus (HDC) score. The HDC score, although promising, relies on the collection of a large number of replicate spectra by the analyst. In this work, we build on the HDC algorithm to construct peak likelihoods . Using a dataset of replicate gas chromatography electron-ionization (EI) spectra of 3 groups of forensically relevant isomers, we show that this peak likelihood score reduces the number of analyte spectra required while increasing classification accuracy. It is well known that EI mass spectral measurements degrade at lower solute concentrations. We quantify this effect on our classification scores, in addition to examining the consequences of using spectra with relative versus absolute intensities.