A method of characterising a sample from spectrometric data using calculation of spectral distance values is disclosed, for use in the field of mass spectrometry. Molecular formula assignment of peaks in mass spectral data is difficult and time-consuming, and the invention provides a computer implemented method of finding a most likely elemental composition of a measured spectral peak of interest. The method analyses isotopic peaks in a portion of the spectrum, using both their mass positions and intensities, to determine a spectral distance between those peaks and isotopic peaks of a candidate composition, finding peaks that match (140). A pattern spectral distance is determined (150) to provide a measure of the correspondence between a set of those peaks in the measured spectrum and peaks of each of a number of candidate compositions. The spectral fit is used to determine a most likely candidate composition.