ANALYSIS OF DIFFERENTIATION STATE IN STREPTOMYCES-ALBIDOFLAVUS SMF301BY THE COMBINATION OF PYROLYSIS MASS-SPECTROMETRY AND NEURAL NETWORKS

The morphological differentiation of Streptomyces albidoflavus SMF301 in a batch culture was analyzed by pyrolysis-mass spectrometry (PyMS). Curie point pyrolysis mass spectra of whole cells at various growth p hases were obtained. The PyMS spectra varied with growth phase and dif ferentiation. It was possible to distinguish differentiation states wi th multivariate statistics and artificial neural network (ANN). Artifi cial neural networks (ANNs) were trained on PyMS data to predict the d ifferentiation state using two different algorithms: backpropagation a nd a radial basis function classifier. Both the backpropagation and th e radial-basis classifier succeeded in separating the differentiation state and identified the transient state. (C) 1998 Elsevier Science B. V. All rights reserved.