Analyzing chromatographic data using multilevel modeling.

摘要: It is relatively easy to collect chromatographic measurements for a large number of analytes, especially with gradient methods coupled mass spectrometry detection. Such data often have hierarchical or clustered structure. For example, analytes similar hydrophobicity and dissociation constant tend be more alike in their retention than randomly chosen set analytes. Multilevel models recognize the existence such structures by assigning model each parameter, its parameters also estimated from data. In this work, multilevel proposed describe time obtained series wide linear organic modifier gradients different duration mobile phase pH acids bases. The consists (1) same deterministic equation describing relationship between analyte-specific instrument-specific parameters, (2) covariance relationships relating various physicochemical properties analyte chromatographically specific through quantitative structure–retention based equations, (3) stochastic components intra-analyte interanalyte variability. was implemented Stan, which provides full Bayesian inference continuous-variable Markov chain Monte Carlo methods.