1. Your distinguished career is rooted in studies of lipoprotein metabolism, particularly the molecular mechanisms of cardiometabolic diseases. What excites you most about the opportunity to directly integrate this deep clinical and molecular knowledge with MedInsights' AI-driven drug repositioning platform, Nexus-Rx?

Cardiometabolic diseases are multiorgan and multimodal in their nature, in which clinical comorbidities are often the result of a complex and heterogeneous biology. MedInsights’ NEXUS-Rx lets us encode that biology directly into an explainable AI framework, instead of treating it like noise. That means moving from correlations to mechanism-anchored predictions and patient stratifications that clinicians can trust. Recent work shows that multimodal AI can materially improve phenotyping and risk prediction in this space, which is exactly where I want my lab’s knowledge to go.

​

2. You have been a vocal proponent of how AI and Network Medicine can 'redefine medicine' and assist clinicians. In rare and metabo-inflammatory diseases, how does MedInsights' approach accelerate the path from lab data to a validated, effective therapeutic option?

Network Medicine treats disease as a disturbance of interconnected modules—not a single faulty molecule—and AI lets us read those modules at scale. When we fuse multi-omics, clinical phenotypes, and exposome data into mechanistic networks, we can pinpoint the endophenotypes that truly drive disease and design therapies against them, rather than just treating downstream symptoms. That shift is the engine of speed and precision. For rare and metabolic-inflammatory conditions, data is sparse and heterogeneous. MedInsights’ network- and knowledge-graph-centric approach aligns with what we’ve published: use heterogeneous networks to integrate diverse evidence (chemistry, targets, pathways, phenotypes), then apply graph learning to prioritize targets and repurposing candidates with explainable rationales. Heterogeneous networks outperform single-source models and weighting schemes and graph methods (e.g., deepDTnet, NeoDTI) improve drug-target predictions, exactly the kind of tool you need when cohorts are small. Graph representation learning enables disease-module discovery, cross-context transfer, and uncertainty-aware, iterative experiment design while keeping models interpretable and resistant to shortcut learning. That means we can move from a computational hit to biological plausibility to patient-level stratification with fewer dead ends. Finally, translation must be human-relevant from day one. In our work, we couple patient-derived multi-omics with organ-specific interactomes and then verify predicted effects in human cell systems - shrinking dimensionality and surfacing testable mechanisms before costly trials. A MedInsights pipeline that operationalizes this “module → mechanism → model → measured benefit” loop will de-risk decisions and accelerate a credible therapeutic to the clinic.

​

3. Moving research from academia into successful clinical application often presents significant challenges. What specific scientific or strategic blind spots are you hoping to help MedInsights avoid, ensuring the Nexus-Rx platform delivers robust, clinically relevant results for patients?

I will always strive after a trustworthy, explainable, healthcare professional and patient friendly platform, which will also inform on economically sustainable and relevant actions. This will ensure the successful implementation of the platform we are developing

​

4. MedInsights focuses on rare and metabolic-inflammatory diseases. Given your expertise in complex metabolic pathways, is there a particular area within these conditions where you believe the combination of your clinical experience and MedInsights’ AI holds the most transformative potential?

MedInsights unifies heterogeneous evidence into an explainable knowledge graph, using graph learning to discover disease modules, predict drug–target links, and prioritize bile-acid transporters/receptors and pathway crosstalk—then rapidly iterates from in-silico hits to patient-level stratification, even with small cohorts. With cross-context transfer and an experiment-guiding loop, it compresses the journey from mechanistic hypothesis to validated therapy and clearly defined responders, shifting care from symptom control to mechanism-based treatment.

​

5. What was the critical factor—whether it was the science, the team, or the mission—that convinced you this was the right time to formalize this strategic partnership with MedInsights?

All the three at the same time. Without science it would be unethical. Without a team it would be undoable. Without a mission it will have no sense

​

​