Strategy Of Metabolic Confounding Effect Elimination And Its Application In Disease Metabolic Profiling Research

ObjectiveConstruct and improve a series of strategies to eliminate the effect of confounding factors in metabolomics data sets,and develop a special software.Apply these strategies to the serum metabolic profiling analysis of typical digestive system cancers for type-and location-specific metabolic signatures identification MethodsBased on independent component analysis,generalized linear model and canonical correlation analysis,a series of strategies was developed for different types of confounding factors and metabolomics data.Their effectiveness was verified by multiple simulation data sets and real data sets derived from different platforms and different types of biosepcimens and diseases.A special software based on Matlab platform was developed.They were applied to eliminate the potential effect of specified confounding factors in serum metabolic profiles of patients with gastric cancer and colorectal cancer and normal controls,from ultra-high performance liquid chromatography quadrupole time of flight mass spectrometry and gas chromatography time of flight mass spectrometry platforms.The characteristic metabolites of different cancer types and locations were identified.A regression diagnosis model was contructed.ResultsThe newly designed strategy can effectively eliminate the influence of different types of confounding factors.The developed software(MCEE V1.0,V2.0)is simple and practical.All the source code,manuals,and sample data were available on Github.Two metabolic markers,taurine and taurine,were found to be related to the type and anatomical position of different digestive tract cancers.The areas under the receiver operating characteristic curves of the anatomical position diagnosis model,which were constructed by 10 characteristic metabolites,were higher than 0.89.ConclusionsCompared with the traditional methods for confounding factor influence elimination or control,the new strategy is stable,robust,and flexible.It does not change the number of samples and metabolites and is compatible to subsequent data analysis.Different strategies could meet the processing requirements of different types and quantities of confounding variables.The software MCEE can serve as one of the useful tools for metabolomics data preprocessing.The characteristic metabolites provided auxiliary diagnostic information,and valuable insights for the research on the commonality of tumor metabolism in digestive system.