Multivariate adaptive embedding (MAE) for the identification of bacterial pathogens in the field

In this work, we demonstrate a new classification machine based on multivariate adaptive embedding (MAE) that is capable of a robust identification of potential bacterial biological warfare agents (BWA). By employing Raman spectroscopy, this method proves to be reliable in application, easy to use and while retaining spectral quality, it is much faster than the often used support vector machines (SVM) and other supervised multivariate statistical classification machines. The multivariate adaptive embedding multi-species classification ability was developed in order to serve as a real-time detection method for biological threat detection and pathogen identification. A mean classification accuracy of 99.25 +/- 0.45% could be achieved with a representative set of biological warfare agents and simulant bacteria as a first approach for a user-friendly and fieldable classification application for first responders and researchers.