A Comparative Study on the Survival Prognosis Model of Immune-related Genes in Patients with Lung Adenocarcinoma based on Deep Learning
10.11783/j.issn.1002-3674.2025.01.010
- VernacularTitle:基于深度学习构建肺腺癌患者免疫相关基因生存预后模型的比较研究
- Author:
Yue REN
1
;
Yang QIN
;
Ning LAN
Author Information
1. 山西医科大学公共卫生学院卫生统计教研室(030001)
- Publication Type:Journal Article
- Keywords:
DeepOmix;
Deep learning;
Prognosis prediction;
Survival analysis;
Lung adenocarcinoma
- From:
Chinese Journal of Health Statistics
2025;42(1):56-61
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the prognostic value of different deep learning models constructed using multi-omics data and immune-related genes in patients with lung adenocarcinoma and compare their predictive performance.Methods DeepOmix,Nnet-survival,Cox-nnet and Deepsurv prognosis prediction models were constructed from the TCGA database lung adenocarcinoma multi-omics data after extracting immune genes using original data and the data after dimensionality reduction by single-factor Cox regression,variance and elastic net(EN)method,respectively.C-index and time-dependent ROC were used to evaluate the predictive effect of the models.Results The results of comparing the C-index,3-year AUC and 5-year AUC values of the multi-omics models showed that the DeepOmix model combined with biological signaling pathway information using different dimensionality reduction methods to construct prognosis prediction models for lung adenocarcinoma had the best prediction performance compared with the DeepOmix,Nnet-survival,Cox-nnet and Deepsurv models(C-index was above 0.83,3-year AUC was above 0.89 and 5-year AUC was above 0.94).In screening variables to construct prognosis prediction models,EN had good predictive accuracy in the majority of cases.The DeepOmix model significantly distinguished between patients in the high-risk and low-risk groups,and the prognosis was worse in the high-risk group(P＜0.001).Conclusion The DeepOmix model combined with biological signaling pathway and EN dimension reduction can analyze high-dimensional low sample size data,and construct a prognosis prediction model for lung adenocarcinoma with high predictive performance compared with other models.
