A machine learning-random forest model has been developed to construct a gene signature associated with cuproptosis aimed at predicting the prognosis of gastric cancer (GC). This advancement is particularly significant as gastric cancer ranks as the fifth most common cancer globally and the third leading cause of cancer-related deaths.
The study led by researchers from Gansu Provincial Hospital and utilizing comprehensive data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) presents new insights for the diagnosis and treatment of gastric cancer, which often progresses undetected until late stages.
Using transcriptome data and diverse computational methods, the researchers created a prognostic model comprised of cuproptosis-related genes (CRGs). After rigorous analysis, seven predictive genes—RTKN2, INO80B, EFNA4, ELF2, MUSTN, KRTAP4, and ARHGEF40—were identified as significant indicators of patient outcomes. These genes showcase how cuproptosis, a novel type of cell death linked to copper toxicity, plays a role in cancer prognosis.
The research highlights the complexity of gastric cancer, where most patients are diagnosed at advanced stages. While early-stage GC can have over 90% 5-year survival rates following surgery, only 25% of internal cases lead to long-term survival due to late detection. Hence, enhancing prognostic models is urgent and necessary for improving patient outcomes.
The model leverages machine learning techniques, particularly the random forest algorithm, to handle large datasets and mitigate risks associated with overfitting. By integrating multiple decision trees, the team improved the accuracy of their predictions about gastric cancer prognosis.
Analysis of the data indicated different pathways dominating the high-risk patient group, predominantly linked to ANGIOGENESIS and TGF_BETA_SIGNALING. This relationship points toward the biological mechanisms at play during cancer progression.
Importantly, the study also assessed the expression of EFNA4, one of the identified genes, through immunohistochemical analysis. Results showed EFNA4 expression significantly elevated compared to normal gastric tissues, with its expression correlates indicating even improved prognosis for patients with higher levels. This suggests EFNA4 might serve as both a prognostic marker and potential therapeutic target.
Given these insights, the prognostic model based on CRGs is expected not only to assist medical professionals but also to lead to personalized treatment approaches for gastric cancer patients. By constructing nomograms, clinicians can quantitatively determine these patients’ potential outcomes—advancing toward more effective treatment plans.
Future discussions will focus on the biological functions of these genes and how to leverage their findings to refine prognostic resources and clinical interventions, aiming to effectively address gastric cancer management and treatment at multiple levels.