Identifying reliable biomarkers - molecular signals that correlate with cancer behavior - is essential for improving prognosis and guiding treatment. In endometrial cancer (EC), despite progress in genomic profiling, the functional roles of most disease-associated genes remain poorly understood.
Weighted Gene Co-expression Network Analysis (WGCNA) is a systems biology approach that organizes thousands of genes into modules based on similar expression patterns across patient samples. Genes in the same module often share biological functions. This allows researchers to narrow from tens of thousands of genes down to a manageable set of biologically coherent candidates.
This study combined WGCNA with eight machine learning algorithms - including gradient boosting, LASSO regression, support vector machines, random forests, and neural networks - to identify hub genes consistently selected across methods as important for EC prognosis. Genes selected by multiple independent approaches are more likely to represent genuine biological drivers rather than statistical noise.
The analysis identified six hub genes consistently selected across all eight machine learning methods as predictive of EC outcomes. Among these, the MAL gene (Myelin and Lymphocyte protein) emerged as the most clinically significant, showing the strongest associations with prognosis across multiple analyses.
High MAL expression was significantly associated with worse overall survival (hazard ratio 2.34, p less than 0.001) and worse progression-free survival (hazard ratio 1.85, p less than 0.001) in EC patients. Patients with high MAL tumors had roughly double the mortality risk compared to those with low MAL expression, a clinically meaningful difference.
MAL expression also correlated with immune microenvironment features: high-MAL tumors showed reduced infiltration of anti-tumor immune cells and elevated markers of immunosuppression, suggesting MAL may help tumors evade immune surveillance. This immune angle makes MAL not only a prognostic marker but potentially a target for immune-based therapies.
To move beyond statistical correlations, the researchers performed in vitro functional experiments in EC cell lines. When MAL expression was experimentally increased in cell cultures, tumor cells showed significantly enhanced proliferation (growth rate) and invasion (ability to penetrate surrounding tissue), both hallmarks of aggressive cancer behavior.
When MAL was silenced using RNA interference, the opposite occurred: cells grew more slowly and invaded less effectively. These experiments provide causal evidence that MAL is not merely correlated with worse outcomes but actively contributes to the aggressive phenotype of high-MAL tumors.
Pathway analysis suggested MAL influences tumor behavior through effects on cytoskeletal organization and membrane lipid raft dynamics - cellular structures important for signaling. The immunosuppressive microenvironment effects may occur through MAL's influence on cytokine secretion patterns, though the precise molecular mechanisms require further investigation.
The study used molecular docking - a computational technique that simulates how drug molecules physically interact with target proteins - to screen for existing drugs that might inhibit MAL. The analysis identified Navitoclax as having favorable binding characteristics to the MAL protein.
Navitoclax showed a binding free energy of -7.5 kcal/mol, indicating stable and energetically favorable binding. In molecular dynamics simulations, the Navitoclax-MAL complex remained stable over time, suggesting the drug could maintain sustained inhibition of MAL protein function.
Navitoclax is a BCL-2 family inhibitor already in clinical trials for various cancers as an apoptosis-inducing agent. Repurposing it as a MAL inhibitor in EC would leverage existing safety data, potentially accelerating translation to clinical use compared to developing a new drug from scratch. However, cell-based and animal model validation of Navitoclax-MAL interactions in EC is still needed.
MAL has the potential to serve as a prognostic biomarker measurable by immunohistochemistry - the same technology routinely used in clinical pathology labs. If validated, pathologists could assess MAL protein expression on standard EC biopsies and use it to stratify patients by risk, potentially alongside existing molecular subtyping.
High-MAL patients with poor prognosis and immunosuppressive microenvironments might particularly benefit from strategies that combine MAL inhibition with immunotherapy. Reducing MAL-driven immunosuppression could restore immune cell infiltration and enhance responses to checkpoint inhibitors.
The study's multi-algorithm machine learning approach - using eight different methods and requiring agreement across all of them - represents a methodological strength that reduces the risk of false discovery. The next critical steps are external validation of MAL's prognostic value in independent patient cohorts and preclinical animal model studies confirming Navitoclax efficacy against MAL-expressing EC tumors.