Análisis de Pseudomonas fitopatógenas usando métodos inteligentes de aprendizaje un enfoque general sobre taxonomía y análisis de ácidos grasos dentro del género Pseudomonas

The identification of plant-pathogenic bacteria is often of high importance. In this paper, we evaluate the identification of plant-pathogenic species within the genus Pseudomonas by fatty acid methyl ester (FAME) analysis. Starting from a FAME database, high quality data sets were generated. Two research questions were investigated: can plant-pathogenic Pseudomonas species be discriminated from each other and can the group of plant-pathogenic Pseudomonas species be distinguished from the group of non-plant-pathogenic Pseudomonas species. In a first stage, a principal component analysis was performed to evaluate the variability within the data. Secondly, the machine learning method Random Forests was evaluated for identification purposes. This intelligent method allows to learn from the variability and patterns in the data and to improve the species identification. The principal component analysis of plant-pathogenic species clearly showed overlapping data clouds. A Random Forests model was developed that achieved a species identification performance of 71.1%. Discriminating the group of plant-pathogenic species from the group of non-plant- pathogenic species was more straightforward, given by the Random Forests identification performance of 85.9%. Moreover, it was shown that a statistical relation exists between the fatty acid profiles and plant pathogenesis