Metabolic phenotyping for the classification of coffee trees and the exploration of selection markers.

High-throughput metabolic phenotyping is a challenge, but it provides an alternative and comprehensive access to the rapid and accurate characterization of plants. In addition to the technical issues of obtaining quantitative data of plenty of metabolic traits from numerous samples, a suitable data processing and statistical evaluation strategy must be developed. We present a simple, robust and highly scalable strategy for the comparison of multiple chemical profiles from coffee and tea leaf extracts, based on direct-injection electrospray mass spectrometry (DIESI-MS) and hierarchical cluster analysis (HCA). More than 3500 individual Coffea canephora and Coffea arabica trees from experimental fields in Mexico were sampled and processed using this method. Our strategy permits the classification of trees according to their metabolic fingerprints and the screening for families with desired characteristics, such as extraordinarily high or low caffeine content in their leaves.