RESUMEN
Laser electrospray mass spectrometry (LEMS) coupled with offline multivariate statistical analysis is used to discriminate eight phenotypes from a single plant organ class and to find potential biomarkers. Direct analysis of the molecules from the flower petal is enabled by interfacing intense (10(13) W/cm(2)), nonresonant, femtosecond laser vaporization at ambient pressure with electrospray ionization for postionization of the vaporized analytes. The observed mass spectral signatures allowed for the discrimination of various phenotypes using principal component analysis (PCA) and either linear discriminant analysis (LDA) or K-nearest neighbor (KNN) classifiers. Cross-validation was performed using multiple training sets to evaluate the predictive ability of the classifiers, which showed 93.7% and 96.8% overall accuracies for the LDA and KNN classifiers, respectively. Linear combinations of significant mass spectral features were extracted from the PCA loading plots, demonstrating the capability to discover potential biomarkers from the direct analysis of tissue samples.