Exploring analyte response in an ESI-MS system with different chemometric tools

Analyte response in mass spectrometry with electrospray ionization was studied for a typical single quadrupole LC-MS instrument, as a function of eight instrumental variables and nine physicochemical properties for a variety of small molecules. Due to the high number of factors, three different experimental designs have been applied in order to balance the amount of gathered information and the number of experimental runs. The analyses were made by nonlinear modeling through artificial neural networks. One of the experimental designs (a Plackett–Burman design) involved 20 compounds of different chemical nature, with 12 runs per compound. It allowed finding general instrumental conditions capable of giving enough sensitivity for all the compounds; some relationships among instrumental parameters and physicochemical properties have also been identified. The other two experimental designs (a 73 runs Doehlert design, and a 19 runs fractional factorial design) were aimed at obtaining optimum instrumental sensitivity for one specific compound; they both allowed identifying the RF component of the Q-array and the desolvation line temperature as the most influential instrumental factors and let us detect several associations among instrumental variables. In order to select the most efficient optimization strategy, these two methods were compared. The fractional factorial design, which requires a much smaller number of runs, was found suitable to identify the optimum working zone, with a fair economy of experiments.