A computer-implemented process and apparatus for automatically identifying compounds in a sample mixture. The process involves receiving a a representation of a measured condition of the sample mixture, using the representation of a measured condition of the sample mixture to select a set of reference spectra of compounds suspected to be contained in the sample mixture, from a library of reference spectra, receiving a representation of a test spectrum having peaks associated with compounds therein, the test spectrum being produced from the sample mixture under the measured condition and combining reference spectra from the set of reference spectra to produce a matching composite spectrum having peaks associated with at least some of the suspected compounds, that match peaks in the test spectrum, the compounds associated with the reference spectra that combine to produce the matching spectrum being indicative of the compounds in the sample mixture.