How does metabolomics enhance computational multiomic models?
My background is in genomics, so I have spent a lot of time working with data that describes genetic variation and molecular regulation. Those data types are highly valuable, but metabolomics offers something different: a functional biochemical readout of biological activity. In computational multiomic models, this is especially useful because it helps bridge the gap between upstream molecular variation and downstream biological effects. Metabolomics can reveal whether signals observed in other molecular layers are reflected in pathway activity and system behavior, thereby making integrated models more interpretable and biologically meaningful. It can also improve predictive modeling and biomarker prioritization by adding a layer that is closely tied to phenotype.
What makes integrating metabolomic data uniquely powerful, and what are the challenges?
The advantage of metabolomic data is its proximity to the phenotype. It adds a layer of functional context that makes integrated analysis much more informative, particularly when you are trying to understand which molecular signals are most biologically meaningful.
The challenge is that metabolomics is often less straightforward to anchor computationally than more gene-centric data types. Linking metabolites to genes, pathways, and broader biological processes can be much more complex and context-dependent, making meaningful integration harder. A lot of my work at Metabolon has focused on improving the links between metabolites and other molecular layers, using curated and bespoke approaches that we hope will make it easier for customers to access more actionable biological insight.
How do AI and machine learning contribute to systems-level discovery?
Multiomic datasets are high-dimensional and heterogeneous, making computational methods essential for identifying patterns and relationships across molecular layers. AI and machine learning are especially valuable for uncovering signals that are difficult to detect with simpler approaches, especially when metabolomics is being integrated alongside other data types.
They can support feature prioritization, reveal non-linear associations, and help identify which signals are most biologically meaningful. Their value lies not just in prediction but in making complex multiomic data more interpretable at the systems level. They are also helpful in making analysis more efficient and accessible, particularly when the goal is to support interpretation rather than simply generate output. Used well, they can help extract signal from large, complex datasets and make it easier to work across multiple molecular layers at once.
What scientific insights are only visible when metabolomics is included?
Including metabolomics makes it much easier to see whether upstream molecular signals are reflected in phenotype-relevant biology. You may observe changes in other molecular layers, but those do not always tell you whether there is a measurable downstream effect.
One of the things I find especially interesting is that not every upstream signal translates into a clear functional consequence. Metabolomics can help distinguish which of those signals are most biologically meaningful by showing whether they are reflected in pathway activity, biochemical disruption, and broader system response. That makes it a critical layer for understanding which molecular changes are most relevant to disease and patient outcome.
What does “From Molecules to Medicine” mean to you in the context of multiomic exploration?
To me, “From Molecules to Medicine” is about turning molecular complexity into something that can make a real difference for patients, whether that is a biomarker, a stronger biological hypothesis, or a clearer understanding of disease.
It also means moving beyond one-size-fits-all approaches. A systems-level view of biology helps explain why patients who appear clinically similar can still differ at the molecular level, which is essential for developing more precise and effective care. Metabolomics is an important part of that because its proximity to phenotype helps connect multiomic data to functional biology and, ultimately, to better patient outcomes.
What advice would you give women pursuing interdisciplinary science careers?
One of the best aspects of a scientific career is that you do not need to stay in one area. Much of my early PhD work was lab-based, but I found I didn’t just want to hand the resulting data over to someone else to analyze; I wanted to follow it through, which naturally led me down a more interdisciplinary path. Being open to learning across different areas can give you a much broader perspective and often makes your work more impactful — it’s important to understand you do not have to limit yourself to one part of the process.
