Personalizing healthcare requires accurate predictions of an individual’s health state. One route to doing so involves the use of minimally invasive biofluid collection, coupled with untargeted metabolomics that provides a wealth of data concerning the person’s state. To date, much focus has been on targeted studies and on developing a handful of biomarkers from those metabolites that can be chemically identified and biologically reasoned about. Here, we wish to evaluate the potential of the complete output of untargeted metabolomics data for developing health state profiles using machine learning techniques. Where possible, we reprocess and analyze 37 published metabolomics studies performed across a range of human health states. In general, we find strong predictive performance using simple and interpretable models. We demonstrate that internal to a study, there may exist significant mutual information between various metabolites and the observed health state, across different chromatographic techniques. Additionally, we find that features frequently not analyzed due to falling above classic, but arbitrary, statistical significance values (p &gt; 0.05 for instance) are generally capable of strong predictive performance suggesting a wealth of additional information in untargeted metabolomics data. These results suggest a data-driven approach for advancing and further personalizing healthcare decisions.

Using machine learning to predict human health from biofluid-based metabolomics

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.