Related Experiment Video
Updated: Jan 11, 2026
11:06
Network Pharmacology Prediction and Metabolomics Validation of the Mechanism of Fructus Phyllanthi against Hyperlipidemia
Published on: April 7, 2023
2.7K
Using Machine Learning for the Discovery and Development of Multitarget Flavonoid-Based Functional Products in MASLD
Maksim Kuznetsov1, Evgeniya Klein1, Daria Velina2
1Department of Food Technology and Bioengineering, Plekhanov Russian University of Economics, 36 Stremyanny per., 115054 Moscow, Russia.
Molecules (Basel, Switzerland)
|November 13, 2025
Summary
This study introduces an in silico pipeline for screening nutraceuticals targeting metabolic dysfunction-associated steatotic liver disease (MASLD). The computational approach rationally designs multi-target formulations for MASLD, offering a transferable framework for metabolic conditions.
Area of Science:
- Computational chemistry and drug discovery
- Pharmacology and toxicology
- Nutraceutical science
Background:
- Metabolic dysfunction-associated steatotic liver disease (MASLD) requires multi-target therapies.
- Traditional approaches often focus on single markers, which is insufficient for MASLD.
- Novel computational strategies are needed for developing effective MASLD treatments.
Purpose of the Study:
- To develop and validate a fully in silico nutraceutical screening pipeline for MASLD.
- To integrate molecular prediction, systemic aggregation, and technological design for nutraceutical formulation.
- To generate rational, multi-target nutraceutical formulations for MASLD.
Main Methods:
- Assembled a panel of ten MASLD-relevant targets from proteomic evidence.
- Extracted and standardized bioactivity data from ChEMBL, generating RDKit descriptors.
- Employed stacked ensemble predictive modeling (Random Forest, XGBoost, CatBoost) with isotonic calibration.
- Utilized physiologically based pharmacokinetic (PBPK) modeling for dose and exposure estimations.
- Integrated physicochemical properties to guide formulation and delivery system design.
Main Results:
- Achieved robust predictive model performance (cross-validated ROC-AUC 0.834, test ROC-AUC 0.840).
- Generated three prototype nutraceutical concepts (HepatoBlend, LiverGuard Tea, HDL-Chews) with defined dosing and exposure profiles.
- Demonstrated successful integration of molecular prediction, systemic coverage, and formulation specifications.
- Identified six structural and twelve mechanism-of-action (MOA) clusters ensuring chemical diversity.
Conclusions:
- The in silico pipeline rationally generates multi-target nutraceutical formulations for MASLD.
- This computational framework links molecular predictions with systemic coverage and practical formulation.
- The approach provides a transferable framework for developing nutraceuticals for MASLD and other metabolic conditions.
