TIGER: technical variation elimination for metabolomics data using ensemble learning architecture

Abstract

Large metabolomics datasets inevitably contain unwanted technical variations which can obscure meaningful biological signals and affect how this information is applied to personalized healthcare. Many methods have been developed to handle unwanted variations. However, the underlying assumptions of many existing methods only hold for a few specific scenarios. Some tools remove technical variations with models trained on quality control (QC) samples which may not generalize well on subject samples. Additionally, almost none of the existing methods supports datasets with multiple types of QC samples, which greatly limits their performance and flexibility. To address these issues, a non-parametric method TIGER (Technical variation elImination with ensemble learninG architEctuRe) is developed in this study and released as an R package (https://CRAN.R-project.org/package=TIGERr). TIGER integrates the random forest algorithm into an adaptable ensemble learning architecture. Evaluation results show that TIGER outperforms four popular methods with respect to robustness and reliability on three human cohort datasets constructed with targeted or untargeted metabolomics data. Additionally, a case study aiming to identify age-associated metabolites is performed to illustrate how TIGER can be used for cross-kit adjustment in a longitudinal analysis with experimental data of three time-points generated by different analytical kits. A dynamic website is developed to help evaluate the performance of TIGER and examine the patterns revealed in our longitudinal analysis (https://han-siyu.github.io/TIGER_web/). Overall, TIGER is expected to be a powerful tool for metabolomics data analysis.

malbacR: A Package for Standardized Implementation of Batch Correction Methods for Omics Data

“…The malbacR package is compatible with pmartR, a package designed for the analysis of omics data. , malbacR specifically includes batch effect correction methods appropriate for MS-based lipidomics and metabolomics data; however, due to its compatibility with pmartR, it can also be used for MS-based proteomics or nuclear magnetic resonance data. Batch correction methods included are pareto scaling, power scaling, range scaling, ComBat, EigenMS, NOMIS, QC-RLSC, RUV-random, WaveICA2.0, TIGER, and SERRF . We group the methods according to their approach: scaling, utilization of QC samples, use of internal standards spiked into each sample, and other parameters.…”

Identification of candidate metabolite biomarkers for metabolic syndrome and its five components in population-based human cohorts

“…Serum samples from the KORA F4 study were measured with the AbsoluteIDQ™ p150 kit (BIOCRATES Life Sciences AG, Innsbruck, Austria) for the quanti cation of 163 metabolites [25]. Speci cally, samples were randomly distributed on 38 kit plates, each plate also including three quality control (QC) samples provided by the manufacturer and one zero sample (PBS) in addition to the individual samples [26,27].…”

“…To minimize the technical variations that metabolomics data inevitably contain, metabolite concentrations were adjusted by a non-parametric method TIGER, which is based on an adaptable ensemble learning architecture [27]. In addition, to ensure comparability between different metabolites, their values were natural-log transformed and standardized to have a mean value of 0 and a standard deviation of 1.…”

Precision Medicine Approaches with Metabolomics and Artificial Intelligence

“…The tremendous complexity of metabolomics data (e.g., peak counts compared to samples), as well as missing values, batch effects during quantification, data noise production, and reproducibility, are all important issues. As a result, the metabolomics community is looking to machine learning approaches to overcome these obstacles [ 26 , 27 ].…”