Abstract
Unraveling the complex associations between human phenotypes and molecular pathways can pave the way to improved health and performance, but faces a fundamental challenge: the measurable genes, proteins, and metabolites vastly outnumber the participants in even the largest studies, yielding spurious correlations. To address this, we developed PhenoMol, a bioinformatic framework that integrates comprehensive phenotypic data predictive of outcomes and reduces multi-omic dimensionality using graph theory constrained by prior biological knowledge. This approach generates biologically informed "expression circuits" to identify causal patterns. Applied to a deeply characterized healthy cohort, PhenoMol successfully predicted elite physical performance and outperformed regression models lacking network-based dimensionality reduction. Designed to be versatile and generalizable, PhenoMol enables studies across small and large populations to predict wellness, performance, and disease outcomes. The software is openly available to support future research in health, disease, and performance optimization.
Introduction
Modern high-throughput biology allows for the measurement of tens of thousands of molecular features—including mRNA, proteins, and metabolites—from a single biological sample. However, human performance and clinical studies often involve relatively small cohorts, typically numbering in the dozens or hundreds. This creates the "curse of dimensionality" ($N \ll P$), where the number of features vastly exceeds the number of subjects, leading to overfitting and the identification of spurious correlations. To bridge the gap between massive multi-omic datasets and meaningful phenotypic outcomes, we developed PhenoMol, a novel bioinformatic framework designed to extract biologically relevant signals from high-dimensional data.
The PhenoMol Framework
PhenoMol addresses the dimensionality problem by constraining data reduction through the use of prior biological knowledge. Rather than treating molecular features as independent variables, the framework maps them onto a comprehensive "interactome"—a large-scale graph representing known protein-protein and protein-metabolite interactions.
The pipeline utilizes the Prize-Collecting Steiner Forest (PCSF) algorithm. Features that correlate with a target phenotype (e.g., fitness scores) are assigned "prizes." The algorithm then identifies the most efficient subnetwork that connects these prized features, often incorporating unmeasured "Steiner nodes" that serve as critical biological bridges. This resulting "Principal Network" is then partitioned into functional modules using Louvain clustering. Finally, Sparse Partial Least Squares Regression (sPLSR) is applied to these biologically grounded modules to build predictive models.
Experimental Design and Data Integration
The framework was validated using a cohort of 86 highly active cadets (71 male, 15 female) from the U.S. Military Academy at West Point. The study integrated 147 phenotypic features—including Army Combat Fitness Test (ACFT) scores, VO2max, body composition, and sleep/activity tracking—with approximately 50,011 molecular features. Multi-omic data included DNA methylation, 3' mRNA sequencing, untargeted proteomics, and metabolomics collected before and after intense physical exertion.
Results and Biological Insights
PhenoMol demonstrated superior predictive accuracy compared to standard statistical methods. For the male cohort, the ensemble models predicted ACFT Total Points with a significantly lower Root Mean Square Error (RMSE) than control models that lacked network-based constraints ($p = 1 \times 10^{-7}$).
Beyond prediction, the framework identified key biological pathways driving elite performance. Over-representation analysis of the high-performing network modules highlighted:
- Complement and Coagulation Cascades: Suggesting the importance of immune and healing responses in physical resilience.
- Primary Bile Acid Biosynthesis: Indicating a central role for energy metabolism and lipid signaling.
- Arginine and Proline Metabolism: Including the urea cycle, which is essential for clearing exercise-induced ammonia and delaying fatigue.
Conclusion
By leveraging established biological architecture as a constraint, PhenoMol effectively filters noise and identifies the molecular "expression circuits" underlying complex phenotypes. While the current study was limited by a small female sample size and a "ceiling effect" among elite performers, the framework's success in predicting fitness outcomes demonstrates its potential for broader applications. PhenoMol is an open-source tool that can be applied to diverse research areas, from rare disease characterization to the optimization of human performance and wellness.
Summary updated