Abstract: |
Background: Deficiencies in nutrients are known to contribute to DNA damage and elevate cancer risk. Standard treatments for lung cancer include surgery, radiation therapy, and chemotherapy. While these drugs are effective in combating lung cancer, they come with considerable side effects, including nausea, vomiting, and hair loss. Optimizing dietary intake of vitamins and minerals could play a role in cancer prevention. However, the specific oncogenic pathways influenced by nutrient deficiencies are not well understood. This study introduces an evolutionary learning -derived method, EL-CoxPH, aimed at uncovering the roles of nutrients in the oncogenic processes of lung adenocarcinoma.
Method: We obtained the gene expression profiles and corresponding clinical data from 493 patients from the TCGA-LUAD dataset of The Cancer Genome Atlas. Each profile included 1,535 miRNAs along with survival-related information, such as patient status and survival time. EL-CoxPH constructs a molecular nutrition network by leveraging an inheritable bi-objective combinatorial genetic algorithm (IBCGA). EL-CoxPH identifies a prognostic miRNA signature and develop a Cox proportional hazards model. The framework encompasses three key identification tasks: discovering a novel survival-associated miRNA signature with prognostic relevance, identifying hub genes targeted by this signature, and exploring pathways as well as nutrient and drug associations related to these hub genes.
Results: EL-CoxPH identified a 15-miRNA signature and established a Cox proportional hazards model, achieving a concordance index (C-index) of 0.654. This performance surpassed that of established prognostic models, including Cox-LASSO, Cox-Elastic, and random survival forest, demonstrating the effectiveness of the proposed method in survival prediction. The identified 15-miRNA signature, ranked in descending order based on their main effect difference scores, included hsa-mir-582, hsa-mir-4661, hsa-mir-212, hsa-mir-4443, hsa-mir-29b-2, hsa-mir-548b, hsa-mir-29b-1, hsa-mir-5571, hsa-mir-132, hsa-mir-142, hsa-mir-584, hsa-mir-3130-2, hsa-mir-374a, hsa-mir-30e, and hsa-mir-548q. In addition to identifying prognostic miRNAs, EL-CoxPH identified six key nutrients—curcumin, quercetin, resveratrol, selenium, vitamin E, and zinc. These findings underscore the potential role of nutrient deficiencies in cancer progression, particularly through their regulatory impact on oncogenic pathways.
Conclusion: This study introduces EL-CoxPH, an evolutionary learning-derived method that integrates machine learning and systems biology approaches to enhance the analysis of oncogenic mechanisms associated with nutrient deficiency. By constructing a molecular nutrition network, EL-CoxPH identifies key nutrients with potential protective effects and elucidates the oncogenic pathways influenced by nutrient deficiencies. Using lung adenocarcinoma as a representative model, the proposed method provides novel insights into the complex interplay between nutrient availability and cancer progression, offering a foundation for further research into targeted nutritional interventions in oncology. |