Identification analysis of ATGs
The expression values of 221 ATGs were integrated from the female patients with lung adenocarcinoma, and 12 downregulated genes and 15 upregulated genes were identified. Boxplots (Figure 1a), Heat maps (Figure 1b), and volcano maps (Figure 1c) were used to reveal the expression patterns of these differentially expressed genes in tumor and non-tumor tissues.
Functional enrichment analysis
A total of 27 differentially expressed genes were performed functional analysis to provide biological function for these genes. In order to inquire the potential signal pathways related 27 ATGs in female lung adenocarcinoma, we analyzed them with GO and KEGG. GO analysis (Figure 2a-c)showed that these ATGs were enriched in some basic biological processes (BP), including the process of autophagy and apoptotic signaling pathways. The cellular components affected those of autophagy and mitochondria. Based on molecular functions, the genes were most enriched in cellular response to oxidative stress, protein phosphatase binding, and phosphatase binding. KEGG analysis (Figure 3a,b)showed that the 27 ATGs were significantly associated with autophagy and mitochondria (animals), and the most scores of enriched pathways were greater than zero, indicating that these pathways were more likely to be enhanced.
Construction of prognostic markers in female LUAD
28 ATGs were analyzed by univariate Cox regression. 8 genes were identified as protective factors (HR<1), while another 20 genes were identified as risk factors (HR>1). The 28 genes were further analyzed by multivariate Cox regression analysis, and finally 12 genes ( ITGA6、ERO1A、FKBP1A、BAK1、CCR2、FADD、EDEM1、ATG10、ATG4A、DLC1、VAMP7、ST13) were identified as independent prognostic indicators, which might be helpful in predicting prognosis.
The 12 genes were further analyzed by multivariate Cox regression analysis, the expression coefficients of each independent risk gene were obtained and shown in table 1, and finally the model for predicting prognosis based was developed using the following formulas: prognostic index (PI) = (-0.162*ITGA6 expression level) + (0.466*ERO1A expression level) + (0.872*FKBP1A expression level) + (0.373*BAK1 expression level) + (-0.392*CCR2 expression level) + (0.416*The expression level of FADD) + (-0.625*EDEM1) + (0.778*ATG1 0 expression level) + (-0.899*ATG4A expression level) + (0.399*DLC1 expression level) + (0.367*VAMP7 expression level) + (0.572*ST13 expression level).
The risk score of each patient was calculated on the basis of the relevant mRNA expression level and risk coefficient of each ATG. The risk score is used to predict the prognosis of female LUAD, and the median risk score is the critical value to divide patients into high-risk and low-risk groups (Figure 5b). Heatmap was drawn to show gene expression profiles in high-risk and low-risk female LUAD groups (Figure 5c). The genes with HR>1 were considered to be dangerous genes, while the genes with HR<1 were the protective genes (Figure 4b). As shown in the figure 4, patients in the high-risk group have more possibilities to express the risk genes. In contrast, patients in the low-risk group have a disposition to express the protective genes(Figure 4b), and the Kaplan–Meier cumulative curve showed that the survival time of patients with low-risk score was significantly longer than that of patients with high-risk score at 10-year follow-up (p < 0.05) (Figure 4a; Figure 5a).
Autophagy acted as an independent prognostic factor
We assessed the prognostic value of autophagy-related risk score by univariate COX regression analysis and multivariate COX regression analysis, the autophagy-related risk score in univariate analysis was significantly correlated with overall survival (OS) (HR=1.235,95% CI=1.173-1.299, P<0.001) in T, M, N stage(Figure 6a). Multivariate analysis further showed that autophagy-related risk score was an independent prognostic indicator(HR=1.208,95% CI=1.143-1.277)(Figure 6b).The results confirmed that the autophagy-related risk score could be used as an independent prognostic factor in clinic practice.
Multi-index ROC curve of risk score and other indicators
The ROC curve of OS has applied for revealing the predictive performance of the 12 autophagy-related gene risk scores (Figure 6c). The AUC value of risk score (AUC=0.842) was significantly larger than that of other indicators including age (AUC=0.510), tumor stage (AUC=0..755), T stage (AUC=0.682), N stage (AUC=0.662) and M stage (AUC=0.578), suggesting that the risk would better predict survival in female patients with LUAD than other individual indicators. However, due to the lack of clinical data such as treatment regimen, personal history, and tumor grade, we were unable to perform ROC analysis for other clinical factors.
Association between autophagy-related risk characteristics and clinicopathological characteristics in female patients with LUAD
These related genes were differentially expressed among various clinicopathological parameters. As shown in the figure 7, different ATG4A expression was found in different tumor stages (Figure 7c), and different BAK1 expression was found in different M stages (Figure 7e). Different CCR2 expression was found in different tumor stages (Figure 7g), different DLC1 expression was found in different tumor stages and N stages (Figure 7b, i), different ERO1A expression was found in different tumor stages and M stages (Figure 7d, f), different FKBP1A expression was found in different N stages (Figure 7h), and different ITGA6 expression was found in different age stages (Figure 7j). Different risk score expression was found in different tumor stages (Figure 7a).