Machine learning based approaches using lipidomics in preclinical and clinical studies using the R Statistical Language.

Statistical analysis of high-dimensional lipidomics data in clinical and preclinical research requires flexibility often absent in proprietary software. The R Statistical Language, which is freely accessible and widely used, provides a variety of packages that can accommodate a variety of statistical approaches necessary to relate lipid biomarkers to clinical and preclinical parameters. In particular, predictive modeling algorithms used in machine learning applications are accessible in the R environment and can be easily applied to lipidomics outputs. In fact, there are several R packages that conveniently supply the modeling architecture to handle all aspects of predictive modeling. This talk will highlight how a little bit of coding can connect researchers to powerful predictive algorithms using the R caret package. The caret package has a streamlined coding workflow for over 200 predictive algorithms across different R packages, and has support functions for data pre-processing, feature selection, model tuning, data splitting, and much more. Examples of this workflow will be demonstrated via biomarker identification studies utilizing untargeted complex lipidomics assays in clinical and preclinical studies. By showcasing user-defined workflows in R, this talk aims to emphasize the power of the R Statistical Language in integrating high-dimensional lipidomics data with diverse datasets, revealing intricate connections between lipid metabolism and clinically measured health outcomes.