Multi-stage search for microbe mass spectra in reference libraries

摘要

The invention relates to the identification of microbes in a sample by calculating the similarities between a mass spectrum of the sample and reference mass spectra in large spectral libraries. The invention provides a multi-stage method for faster identification because the similarity analysis for a large library of 10,000 reference spectra takes around one minute with today's computers, while the spectrum acquisition itself takes only a few seconds. In this multi-stage method, the search is initially confined to a relatively small sub-library of a first stage, containing only the reference spectra of a few microbes which, statistically, are found particularly often and, in addition, all reference spectra that have at least a minimum level of similarity to the reference spectra of these microbes. In routine laboratories, it is often the case that 50 percent of all routine analyses relate to only five species of microbe, for example; the sub-library which contains all reference spectra that are sufficiently similar to the spectra of these species of microbe contains around one hundred reference spectra, and the search in this sub-library of the first stage takes much less than one second. Only if the search does not provide a hit with sufficient certainty of identification, is the search carried out in sub-libraries of higher stages compiled in an analogous manner or, finally, in the whole library. A hierarchy of sub-libraries can be generated automatically, and it is possible to minimize the average computing time for the identification by taking into account the latest statistics and by selecting right numbers and sizes of the sub-libraries.