Discriminant Function Analysis Based on Principal Components for Rapid Discrimination of Metabolic Capabilities of New Isolates

Abstract

The growing need for new microorganisms with novel metabolic capabilities has urged scientists to search for life in extreme environments. With the rapid progress in experimental methods, it is possible to isolate new microorganisms at high speeds but the bottleneck in this process is the biochemical characterization due to time and financial limitations. Inferential hierarchical clustering of new isolates may help to overcome this problem. In this work, discriminant function analysis, used in conjunction with principal component analysis (PCA) was able to rapidly discriminate eight new isolates using metabolic footprints (spectral data from electrospray injection mass spectrometry) and the results were compared with clustering based on the Euclidean distances computed both in the domain of original data and in the domain of PCA-transformed data. The presence of the replicates on the adjacent leaf nodes of dendrograms obtained by hierarchical cluster analysis confirmed the reliability of the method. This attractive tool is applicable to a chemical/biological system, which involves complex samples that can be analyzed by high-throughput instruments.