Machine learning data mining technique finds new potential interactions between drugs and supplements

The new research was published this week in the Journal of Biomedical Informatics​. The authors are associated with universities and institutions in the United States and Brazil.

The authors’ stated goal was to improve understanding of the terminology used in the study of food ingredients and finished supplement formulations with a view to matching those with plausible modes of drug-supplement interactions. They did this in “by leveraging biomedical natural language processing (NLP) technologies and DS (dietary supplement) domain terminology.”

Teaching new tricks to an old tool

The researchers started with an NLP tool called SemRep​​. which is a “UMLS-based program that extracts three-part propositions, called semantic predications, from sentences in a biomedical text.”

UMLS, or Unified Medical Language System​, “Integrates and distributes key terminology, classification and coding standards, and related resources to promote the creation of more efficient and interoperable biomedical information systems and services, including electronic health records.

The nomenclature of food ingredients has been problematic for decades, especially for botanical ingredients. A key part of the effort was getting the terminology right, so the new NLP tool the researchers developed knew what to look for and reduced false feedback.

Hundreds of new potential supplement-drug interactions discovered

They called their new tool SemRepDS, which added new dietary supplement terminology to the basic SemRep tool. They added over 28,000 supplement-specific terms to create their new tool. When applying both tools to studies accessible via PubMed, SemRepDS “Returned 158.5% more DS entities and 206.9% more DS relationships than SemRep.”

The researchers found 350 potential drug-supplement interactions, 325 of which did not appear to have been mentioned in the literature before. After review by experts in the field, 76 of these novel interactions were judged to have plausible mechanisms of action.

The results of the data mining were plotted on a comprehensive knowledge graph, which the researchers called SuppKG.

Kingston: New approach shows promise, but some underlying data is weak

Rick Kingston, PharmD, scientific and regulatory affairs manager for SafetyCall International, said the new research is intriguing. But he said the problematic nomenclature of food ingredients goes hand in hand with deeper difficulties with much of the research. Although this is less true today, many studies performed on food ingredients in the past may have suffered from poor ingredient characterization, making it difficult to use this research in a unified data set. .

“This modeling appears to be a novel approach to identifying potential drug-supplement interactions. Although the model may be applicable in a variety of settings, there are likely important limitations when it comes to dietary supplements. The model assumes that the characterization and identification of dietary supplements found in published work is accurate,”said Kingston (who is also a professor at the University of Minnesota).

“Although the characterization of various botanicals has improved, historically this level of detail has been lacking in many published studies. Another problem is that many dietary supplements on the market contain multiple ingredients in a proprietary blend that does not include the exact concentration of the individual ingredients. Thus, there may be effects resulting from the combination of multiple ingredients rather than just one ingredient in the mix,”he added.

Despite its limitations, this modeling can generate reasonable assumptions regarding the ingredients of single-substance dietary supplements that can contribute to the understanding of supplement-drug interactions. The next step is to investigate some of the potential interactions and determine if they have clinical relevance,”concluded Kingston.

Source:Journal of Biomedical Informatics
2022 Jun 13;104120. doi: 10.1016/j.jbi.2022.104120. Online ahead of print.
Discover new drug-supplement interactions using SuppKG generated from biomedical literature
Authors: Schutte D, et. Al.

Sherry J. Basler