This large-scale study by the Clinical Proteomic Tumor Analysis Consortium (CPTAC) analyzed 138 endometrial cancer tumors using 10 different molecular analysis platforms simultaneously, including genomics, proteomics, phosphoproteomics, and glycoproteomics. The result is the most detailed molecular portrait of endometrial cancer assembled to date.
By stacking these multiple layers of molecular data, the researchers could move beyond simply cataloguing which genes are mutated to understanding how those mutations affect protein function and downstream signaling in real cancer tissue, revealing new opportunities for targeted therapies.
Genomics identifies which genes are mutated or amplified. Proteomics measures which proteins are actually produced from those genes. Phosphoproteomics tracks which proteins are activated or deactivated through phosphorylation (a chemical modification that acts like an on/off switch). Glycoproteomics examines sugar modifications on proteins that affect cell communication.
The power of combining all these layers lies in resolving discrepancies: a gene may be mutated in DNA but the resulting protein may be compensated for by other pathways. Only by measuring proteins and their modifications can researchers determine which genetic changes actually matter for tumor biology.
One of the study's most actionable findings involves PIK3R1, a gene involved in regulating the PI3K/AKT signaling pathway, which controls cell growth and survival. The researchers found that specific types of PIK3R1 mutations (called in-frame indels) are associated with increased activation of the AKT protein.
Critically, these mutations appear to predict sensitivity to AKT inhibitors, a class of targeted drugs already in clinical development. This means patients whose tumors carry PIK3R1 in-frame indels could potentially be selected for AKT inhibitor therapy based on their tumor's molecular profile, a major step toward personalized endometrial cancer treatment.
The study developed a practical SRM (Selected Reaction Monitoring) assay using just two proteins to predict a tumor's antigen presentation machinery (APM) status, which determines whether the immune system can recognize and attack the cancer. The two-protein assay achieved an AUC of 0.961, meaning it could accurately predict APM status in nearly all tumors tested.
Knowing APM status is clinically valuable because tumors with intact antigen presentation tend to respond better to immunotherapy drugs like checkpoint inhibitors. This simple protein test could help oncologists identify which endometrial cancer patients are most likely to benefit from immunotherapy.
The study also trained a deep learning model to predict endometrial cancer molecular subtypes directly from routine H&E pathology slide images, without requiring any genetic testing. The model predicted POLE mutation status (one of four recognized molecular subtypes with an excellent prognosis) with an impressive AUROC of 0.925.
POLE-mutated tumors have the best outcomes of any endometrial cancer subtype and may be candidates for treatment de-escalation. If a slide-reading AI can reliably identify these patients, it could guide treatment decisions in settings where expensive molecular testing is unavailable or cost-prohibitive.
The researchers identified MYC activity (a measure of how active the MYC cancer-driving protein is in a tumor) as a potential biomarker for predicting sensitivity to metformin, a diabetes drug that has shown anti-cancer properties in laboratory studies. Tumors with high MYC activity appeared more sensitive to metformin's growth-inhibiting effects.
Metformin is inexpensive, widely available, and generally well tolerated. If MYC activity can reliably identify endometrial cancer patients who will benefit from metformin, this could open a low-cost therapeutic option, particularly for patients whose tumors have high MYC-driven proliferation.
This comprehensive proteogenomic atlas of endometrial cancer goes far beyond cataloguing mutations to identify druggable proteins and pathways that are actually active in tumors. The multiple layers of molecular data revealed connections between genetic alterations and protein-level consequences that neither genomics nor proteomics alone could provide.
The practical outputs, including the PIK3R1 mutation guide for AKT inhibitor selection, the two-protein APM assay, and the slide-reading AI for POLE detection, are designed to be translatable to clinical use. The dataset itself is publicly available, enabling other researchers to mine it for additional therapeutic insights.