Pancreatic cancer has the poorest prognosis of any major cancer, with a 5-year survival rate of only about 11% overall and just 3% in patients with metastatic disease. The cancer's silent nature means most patients are diagnosed late, when curative surgery is no longer possible. Even patients who do undergo surgery face a 75% recurrence rate.
Finding reliable biomarkers could improve early detection and reveal new therapeutic targets. This study used machine learning to sift through large gene expression datasets from thousands of patients, identifying genes that are consistently altered between cancerous and healthy pancreatic tissue, then validating the most promising ones experimentally.
Researchers analyzed two large gene expression datasets (GSE62452 and GSE28735) from the Gene Expression Omnibus (GEO), comparing pancreatic cancer tissue to normal tissue. They identified 35 consistently differentially expressed genes. Machine learning was then applied to these 35 genes across multiple datasets to narrow the list to eight diagnostic genes involved in cancer progression.
Further filtering using three additional GEO datasets identified three critical genes that were most consistently associated with pancreatic cancer: CTSE, LAMC2, and SLC6A14. A diagnostic model built on these three genes showed strong ability to distinguish cancerous from normal tissue in both training and validation datasets from GEO, TCGA, and independent clinical cohorts.
Analysis of TCGA data revealed that both LAMC2 and SLC6A14 were expressed at higher levels in more advanced clinical stages of pancreatic cancer. Patients with high expression of these genes had significantly worse survival, identifying them as not just diagnostic but also prognostic markers.
Both genes also showed strong correlations with immune cell abundance in the tumor microenvironment. Higher LAMC2 and SLC6A14 expression was associated with more abundant immune cells including macrophages and other populations. This immune connection suggests these genes may contribute to the immunosuppressive environment that helps pancreatic cancer evade the body's defenses.
To understand what SLC6A14 actually does in cancer cells, researchers used laboratory experiments to knock down its expression in pancreatic cancer cell lines. Reducing SLC6A14 significantly suppressed cell proliferation (confirmed by EdU assays and colony formation), migration, and invasion—all key behaviors that drive cancer spread.
Mechanistically, SLC6A14 appears to act through the Wnt/beta-catenin signaling pathway, a well-known cancer-promoting pathway. When SLC6A14 was suppressed, markers of this pathway decreased, and epithelial-to-mesenchymal transition (EMT)—a process that helps cancer cells metastasize—was also inhibited. This confirms SLC6A14 is not merely a bystander but an active driver of cancer progression.
The three-gene diagnostic model (CTSE, LAMC2, SLC6A14) demonstrated strong performance in distinguishing cancer from normal tissue across multiple independent datasets, suggesting it could potentially form the basis of a clinical test. Such a test could help screen high-risk individuals or aid in confirming diagnoses from imaging.
Because SLC6A14 is a transporter protein (it moves amino acids across cell membranes), it is located on the cell surface—making it accessible to antibody-based drugs without needing to enter the cell. This surface accessibility makes it an attractive candidate for targeted therapy development.
By combining machine learning-based gene selection with rigorous laboratory validation, this study identified SLC6A14 as a promising new biomarker and potential therapeutic target in pancreatic cancer. The diagnostic model based on three genes showed consistent performance across multiple datasets, providing confidence in its reliability.
Future research should test whether blocking SLC6A14 in animal models and eventually in patients can slow cancer progression. Clinical trials targeting this pathway could open new treatment options for a disease with very limited therapeutic choices.