Distinguishing analyte from noise components in mass spectra of complex samples: where to cut the noise?

Fourier transform mass spectrometry (FTMS) enables comprehensive analysis of complex molecular mixtures. Given the broad intensity ranges of components in the mass spectra, it is imperative to accurately determine a noise threshold level above which peak assignments will be made. Conventionally, to find the threshold level, the "N sigma" approach or an equivalent rule is used. However, the "N sigma" approach cannot be applied to mass spectra stored with partially removed noise (reduced-profile mode). It is also not directly applicable to mass spectra acquired in the absorption mode with removed negative spectral amplitudes. Moreover, N value selection is normally made based on a rule of thumb, meaning that the calculated threshold level may be biased. Here, we present a noise thresholding method which addresses these limitations for analysis of mass spectra of complex molecular mixtures. The introduced data-dependent thresholding method involves analysis of the distribution of logarithmic intensity of all peaks, including noise and analyte, for a given mass spectrum. Selected method applications include FTMS analysis of crude oil fractions as well as tandem MS analysis of intact proteins.