APPLICATION OF PATTERN-RECOGNITION TO MONITORING FERMENTATIONS OF BACILLUS-AMYLOLIQUEFACIENS

Pattern recognition techniques were applied to analytical data to distinguish abnormal from normal microbial fermentations using Bacillus amyloliquefaciens as a model system. Patterns of fermentation end products during growth of B. amyloliquefaciens were obtained from HPLC analysis of broth samples. Data were also obtained from fermentations using other bacterial species, strains, and environmental conditions, and were compared with the model data set. The bacterial species cultured included B. subtilus, B. licheniformis, and Escherichia coli. Environmental variables included aeration and temperature. The chromatographic patterns were compared by using hierarchical cluster and principal component analysis to obtain a quantitative measure of their similarity and to establish the normal variability within a model data set. Statistical analysis of the data indicated that individual fermentations can be assigned to distinct clusters on the basis of their divergence from the model system. Altered environments and other species can be identified as outliers from the model set. These results show that pattern recognition analysis has direct applicability to monitoring fermentation processes.