Title: Reaching the summit: the importance of feature detection
Feature detection is typically the first step of data processing in liquid chromatography mass spectrometry (LC-MS)-based metabolomic studies. Although downstream statistical analyses rely on precise identification and integration of peaks (features), the importance of this step is often underappreciated in practice. I will present an overview of various existing approaches to feature detection in high-resolution LC-MS data based on extracted ion chromatograms, wavelet transform algorithms, Kalman filters, and others. All methods have their potential strengths and pitfalls, but none of them can produce optimal results without parameter optimization, which should ideally be done systematically using a design-of-experiment approach. Since benchmarking feature detection algorithm is a non-trivial problem by itself, I will outline potential approaches to this problem. Finally, I will discuss possible future developments towards more intelligent methods for feature detection.