MCEE 2.0: more options and enhanced performance

A confounding factor is an unstudied factor that affects one or more of the variables that are being studied in an investigation, so the presence of a confounder may lead to inaccurate or biased results. It is well recognized that physiological and environmental factors such as race, diet, age, gender, blood pressure, and diurnal cycle affect mammalian metabolism. To eliminate the noise introduced by confounders into metabolomics studies, a GUI-based method denoted metabolic confounding effect elimination (MCEE) was developed and has since been applied successfully in a wide range of metabolomics studies. To keep up with recent developments in computational metabolomics and a growing number of user requests, an upgraded version of MCEE with more options and enhanced performance was designed and developed. Besides the generalized linear model (GLM) method, a multivariate method for selecting affected metabolites—canonical correlation analysis (CCA)—was introduced, which accounts for complicated correlations and collinearity within the metabolome. Multiple confounders are acceptable and can be identified and processed separately or simultaneously. The effectiveness of this new version of MCEE as well as the pros and cons of the two methodological options were examined using three simulated data sets (a basic model, a model with different sample size ratios, and a sparse model) and two real-world data sets (a human type 2 diabetes mellitus data set and a human arthritis data set). As well as presenting the results of this examination of the new version of MCEE, some instructions on appropriate method selection and parameter setting are provided here. The freely available MATLAB code for MCEE with a GUI has also been updated accordingly at https://github.com/chentianlu/MCEE-2.0 . Graphical abstract