Key points are not available for this paper at this time.
Abstract The capability of the ‘Self‐Training Interpretive and Retrieval System’ has been extended so that the compounds selected as providing the best match in each data class to the unknown mass spectrum are examined by the computer for the presence of each of 179 common substructural fragments. Stastical methods were used to evaluate the selectivity for identification of each substructure by each data class using a reference file of 18 806 spectra of different compounds. In tests using at least 373 unknown spectra for each of the substructures, with criteria that gave a mean reliability of 98.1%, the method averaged 49% recall, which corresponds to the identification of 2.55 substructures per unknown spectrum, as well as the normal “Self‐Training Interpretative and Retrieval System” match‐factor results, requires 68 s on a laboratory computer. The method is also available to outside users on an international computer network.
Dayringer et al. (Sat,) studied this question.