Screening of FRLs
After obtaining the list of FRGs, the expression levels of the lncRNAs and FRGs were extracted from GSE81089. A total of 61 DE-FRGs and 303 DELs were screened in LUAD and normal tissues. These genes are shown in volcano plots (Fig. 1A) and heatmaps (Fig. 1B).
The relationship between DE-FRGs and DELs was explored. A total of 621 pairs were obtained, and the network is shown in Fig. 1C. Among these, 223 FRLs were used for further analyses.
Construction Of A Prognostic Model Based On Frl Pairs
After the screening, 621 FRL pairs were identified. Among these, 289 lncRNA pairs were associated with the prognosis of patients with LUAD. Furthermore, these pairs were subjected to LASSO analysis to select the optimal combination, and 18 lncRNA pairs were finally obtained to construct the risk model (Fig. 2A). Among these, 11 were protective FRL pairs and 7 were risky FRL pairs (Fig. 2B).
Validation Of The Prognostic Model
As shown in Fig. 3A, the AUROC for the prognostic model was 0.943 and the cut-off value was − 0.045. Based on the threshold, all patients with LUAD were classified into HR (n = 63) and LR (n = 45) groups. The risk curve and scatter plot showed that the number of deceased patients increased with increasing RS (Fig. 3B). Similarly, the K-M curve also indicated that patients in the HR group had significantly worse prognosis than those in the LR group (p < 0.0001, Fig. 3C). The ROC curve showed that the AUC values were 0.976, 0.962, and 0.938 for 1-, 3-, and 5-year time points (Fig. 3D), respectively, indicating that the constructed model had excellent predictive ability for LUAD prognosis.
We also verified the performance of this model using the TCGA data. Deaths were mainly observed in the HR group (Fig. 3E). In addition, a significant difference in the survival rate was observed between the LR and HR groups (p = 0.015, Fig. 3F). The AUC values at 1-, 3-, and 5-year time points were 0.746, 0.745, and 0.737, respectively (Fig. 3G). These findings further confirm that the model had a good predictive effect.
Correction Between Clinical Features And The Two Different Risk Groups
Furthermore, we analyzed the relationship between the risk groups and differential clinical factors (age, sex, pathological tumor-node-metastasis (pTNM) stage, WHO performance status, and smoking history) in the training set. The results showed that the distribution of pTNM stage and WHO performance status were significantly different between the HR and LR groups (Table 1). Most patients in the LR group had pTNM stage 1 (n = 47), and pTNM stage 4 was not observed in this group. Patients in the HR group mainly had stage 1 and stage 3 pTNM (all n = 15, Fig. 4A and 4B). In terms of WHO performance status, the majority samples in the LR group had WHO performance status 1, and a minority of samples in the HR group had status 2 (Fig. 4C and 4D).
Table 1
Comparison of different clinical factors between low risk and high risk groups.
Clinical characteristics | N | RS status group | P value |
Low(N = 63) | High(N = 45) |
Age(years) | | | | |
≤ 60 | 20 | 12 | 8 | 9.980E-01 |
> 60 | 88 | 51 | 37 |
Gender | | | | |
Male | 39 | 21 | 18 | 5.440E-01 |
Female | 69 | 42 | 27 |
pTNM stage | | | | |
1 | 62 | 47 | 15 | 9.344E-05 |
2 | 19 | 7 | 12 |
3 | 24 | 9 | 15 |
4 | 3 | 0 | 3 |
WHO performance status | | | | |
0 | 73 | 47 | 26 | 5.442E-04 |
1 | 25 | 7 | 18 |
2 | 10 | 9 | 1 |
Smoking history | | | | |
Never | 11 | 7 | 4 | 8.479E-01 |
Ex-smoker | 52 | 29 | 23 |
Current smoker | 45 | 27 | 18 |
Bold represents significant. |
Determination Of Independent Prognostic Factors
Independent prognostic factors for patients with LUAD were determined. PTNM stage and RS were significantly associated with patient survival per the two Cox regression analyses, indicating that they were independent prognostic factors (Table 2).
Table 2
Univariate and multivariate cox regression analysis of clinical characteristic factors.
Variables | Univariate analysis | Multivariate analysis |
HR | 95% CI | P value | HR | 95% CI | P value |
Age | 1.007 | 0.972–1.042 | 7.15E-01 | - | - | - |
Gender | 1.231 | 0.696–2.178 | 4.74E-01 | - | - | - |
pTNM stage | 1.973 | 1.479–2.632 | 1.08E-06 | 1.232 | 1.056–1.677 | 1.84E-02 |
WHO performance status | 1.273 | 0.759–2.138 | 3.59E-01 | - | - | - |
Smoking history | 1.038 | 0.682–1.580 | 8.60E-01 | - | - | - |
RS status (High/Low) | 21.100 | 8.882–50.14 | 4.98E-12 | 18.198 | 7.442–44.502 | 2.03E-10 |
Bold represents significant. |
Development Of The Nomogram
PTNM stage and RS were used to generate a nomogram model. Between the two, RS contributed the most to survival probability (Fig. 5A). As shown in Fig. 5B, the C-indices for 1-, 3-, and 5-year time points were 0.706, 0.790, and 0.813, respectively, and favorable consistency was observed between the actual and predicted survival rates.
Differential Expression Analysis Between The Lr And Hr Groups
A total of 379 DEGs were screened in the HR and LR groups, which are depicted in a volcano plot (Figure S2A). These DEGs were significantly enriched in 17 GO BP terms (Figure S2B) and 16 pathways (Figure S2C). In brief, DEGs were enriched in the regulation of vasoconstriction (GO-BP term) and were primarily involved in neuroactive ligand-receptor interaction, axon guidance, and cytokine-cytokine receptor interaction pathways.
The differences in KEGG pathways between LR and HR groups were determined using GSVA analysis. Sixteen different pathways were identified (Fig. 6A). KEGG pathways, such as pyrimidine metabolism, drug metabolism-other enzymes, and starch and sucrose metabolism were more enriched in the HR group than in the LR group; conversely, pathways such as regulation of autophagy, phosphatidylinositol signaling, and GNRH signaling were enriched in the LR group. Furthermore, a network was constructed to show the correlations between KEGG pathways (Fig. 6B). We found that the phosphatidylinositol signaling system was associated with the GNRH signaling pathway, and pyrimidine metabolism was correlated with the drug metabolism-other enzymes.
Association Between Immune Infiltration And Risk Groups
A total of 3, 0, 9, and 2 different immune cells were obtained between the LR and HR groups using the CIBERSORT, TIMER, ssGSEA, and MCPcounter algorithms, respectively (Fig. 7A). We observed that stromal, immune, and ESTIMATE scores were increased in the HR group, and tumor purity was decreased in the LR group. M2 macrophages and neutrophils were more enriched in the HR group, whereas plasma cells were more infiltrated in the LR group.
In terms of immune checkpoint genes, the expression levels of BTLA, CD274, and HAVCR2 significantly increased, whereas that of CTLA4 significantly decreased in the HR group (Fig. 7B).
Analysis Of The Sensitivity To Agents
After searching the GDSC database, 175 small-molecule drugs used for treating patients with LUAD were obtained. The IC50 values of these agents were calculated. The estimated IC50 values of dasatinib (Fig. 8A), gemcitabine (Fig. 8B), paclitaxel (Fig. 8C), and erlotinib (Fig. 8D) significantly decreased in the HR group, indicating that patients from the HR group were more sensitive to them.