Variable Selection in Chemometrics

Since its emergence, chemometrics has been based essentially on factorial methods, whose aim is to extract meaningful directions in the space of the measured variables, called latent variables, or components. This is certainly one of the reasons for the success of chemometrics, which performs an analysis of high-dimensional data without a priori, and, in particular, without being limited to a subset of the measured variables. The subject of variable selection has therefore been less studied in the world of chemometrics than in that of statistics, where variable selection was seen as a means of solving the curse of dimensionality. However, recent developments are bringing the subject of variable selection back into the spotlight. Many spectrometry techniques, such as near-infrared spectrometry, are now mature enough to be used on-line or as portable sensors. The design of simple, efficient, and low-cost sensors relies on the determination of a limited number of wavelengths, thus on the selection of variables. Over the past decades, analytical chemistry techniques have become increasingly resolute, powerful, fast, and diverse. Users are therefore confronted with the processing of larger and more complex data. For example, the measurement of a single sample by a mass spectrometer can generate several GBs of data. The processing of this data still relies heavily on the assessment of experts, who could be usefully assisted by appropriate variable selection techniques. Finally, variable selection is a tool to reveal the most important features of the measured signals. Thus, in fields such as metabolomics, variable selection can constitute a powerful tool to help in the identification of biomarkers.

This research topic aims to collect contributions from scientists working in various disciplines who have common interests in variable selection for analytical chemistry and chemometrics, either at the methodological or application level. Articles can be original methodological research, applications, reviews, or syntheses. The main objective is to demonstrate the value of variable selection in chemometrics and make the most recent advances available to the related scientific communities.