Mining fatty acid databases for detection of novel compounds in aerobic bacteria.

This study examines how the discriminatory power of an automated bacterial whole-cell fatty acid identification system can be significantly enhanced by exploring the vast amounts of information accumulated during 15 years of routine gas chromatographic analysis of the fatty acid content of aerobic bacteria. Construction of a global peak occurrence histogram based upon a large fatty acid database is shown to serve as a highly informative tool for assessing the delineation of the naming windows used during the automatic recognition of fatty acid compounds. Along the lines of this data mining application, it is suggested that several naming windows of the Sherlock MIS TSBA50 peak naming method may need to be re-evaluated in order to fit more closely with the bulk of observed fatty acid profiles. At the same time, the global peak occurrence histogram has put forward the delineation of 32 new peak naming windows, accounting for a 26% increase in the total number of fatty acid features taken into account for bacterial identification. By scrutinizing the relationships between the newly delineated naming windows and the many taxonomic units covered within a proprietary fatty acid database, all new naming windows were proven to correspond with stable features of some specific groups of microorganisms. This latter analysis clearly underscores the impact of incorporating the new fatty acid compounds for improving the resolution of the bacterial identification system and endorses the applicability of knowledge discovery in databases within the field of microbiology.