Construction, validation, and evaluation of a seven ferroptosis-related lncRNAs prognostic signature
The entire set (N = 506) with 439 FRlncRNAs expression data was randomized into the test set (N = 252) and train set (N = 254). In the univariate Cox regression assessment, 22 FRlncRNAs modulated the overall survival of the patients in the train set (Fig. 1a). Lasso regression was used for further analysis to eliminate overfitting lncRNAs, and the 16 lncRNAs we obtained were used for the subsequent multivariate Cox regression analysis (Fig. 1b-d) (concordance index [C-index], 0.75). The ferroptosis-associated lncRNA prognostic biosignature was developed based by summing up the product of each lncRNA expression with its corresponding coefficient in multivariate Cox regression as indicated below: lncRNA biosignature risk score= (0.136× expression of AC005550.2) + (0.240× expression of LINC02381) + (-1.407× expression of AL137782.1) + (0.365× expression of C2orf27A) + (0.235× expression of AC156455.1) + (0.280× expression of AL354993.2 1) + (0.653× expression of AC008760.1). Analysis using the Proportional Hazards Assumption in the Cox model revealed that all the P values > 0.05, implying they conformed to the PH test (Additional file 9).
According to the median value of the risk score, results of the Kaplan-Meier curves demonstrate that the high-risk group has a remarkably dismal overall survival (OS) in contrast with the low-risk group in the train set (P = 2.899E-06), test set (P = 5.314E-03), and entire set (P = 1.1E-06) (Fig. 2a-c). The train set shows three years' OS for patients with high and low-risk group were 60.6% and 90.5%, respectively. The test set is 63.9% and 90.1%, respectively. The entire set is 60.6% and 90.5%, respectively. The AUC of three years dependent ROC for the seven-lncRNA biosignature achieves 0.796, 0.715, and 0.758 respectively in the train set, test set, and entire set (Fig. 2d-f), which demonstrate the good performance of the model in estimating the CRC patients' OS. The mortality rate was higher in patients with high-risk scores relative to those with low-risk scores in the three sets (Fig. 2g-i). The six lnRNAs’ (AC005550.2, LINC02381, C2orf27A, AC156455.1, AL354993.2, AC008760.1) expression of signature were lower in low-risk group compared to the high-risk group in cluster heat map, AL137782.1 oppositely (Fig. 2j-l).
It is worth noting that AC156455.1, and AL354993.2's high expression of this lncRNA signature also has a worse OS than low (Fig. 3). The association of the seven lncRNAs with ferroptosis genes is shown in Fig. 4. In addition, we stratified according to various clinical factors (clinical stage, gender, age, CEA levels, MMR status, postoperative tumor status, perineural invasion, vascular invasion, KRAS mutation, BRAF mutation) and applied the prognostic model to OS detection, which is shown in Fig. 5, the results shown that the signature has good predictive significance for CRC patients in most stratification factors, and part of results are not satisfactory (P > 0.05), which might be due to there are not enough samples in these stratification.
Independent prognostic analysis of the seven ferroptosis-associated lncRNAs signature and its correlation with clinical variables.
The Univariate Cox regression assessment demonstrated that the lncRNA biosignature risk score was evidently correlated with the patients’ OS (hazard ratio HR = 1.224, confidence interval 95%CI = 1.149–1.303, P = 3.69E-10) in train set, (HR = 1.160, 95%CI = 1.016–1.325, P = 0.028) test set, and (HR = 1.179, 95%CI = 1.125–1.235, P = 7.04E-12) entire set (Table 2). Moreover, the multivariate Cox regression analysis demonstrated that the lncRNA biosignature risk score remained independent with OS considering other conventional clinical factors including Lymph-node status, the clinical stage, distant metastasis, and T stage (HR = 1.187, 95% CI = 1.107–1.273, P = 1.39E-06) in the train set, and (HR = 1.122, 95% CI = 1.067–1.180, P = 8.33E-06) in entire set, in spite of P > 0.05 (HR = 1.059, 95% CI = 0.995–1.289, P = 0.059) in test set. Meanwhile, T stage and age were demonstrated as an independent prognostic index. Compared to clinical variables, this signature risk score's ROC curves of three years demonstrate the second-largest AUC value (0.737) (Fig. 6).
Based on the stratification of clinical variables, the correlation between the lncRNAs and clinical variables shows that LINC02381' expression is related to T stage, Lymph-node status, and clinical stage, KRAS mutation, BRAF mutation, and perineural invasion. C2orf27A' expression is associated with T stage, Lymph-node status, clinical stage, KRAS mutation, MMR. AC156455.1' expression is correlated to Lymph-node status. AL354993.2' expression is connected to distant metastasis, Lymph-node status, clinical stage, KRAS mutation. AC008760.1' expression is concerning to Lymph-node status, distant metastasis, clinical stage, KRAS mutation. AL137782.1' expression is linked to KRAS mutation. The lncRNA signature' riskscore is coupled to T stage, Lymph-node status, distant metastasis, clinical stage, and KRAS mutation. (Fig. 7).
Functional enrichment analysis of the seven ferroptosis-related lncRNAs signature.
GSEA analysis is used to discover potential biological functions of the seven ferroptosis-associated lncRNAs signature of CRC (Fig. 8). The results showed that three signaling pathways (KEGG_HEDGEHOG_SIGNALING_PATHWAY, KEGG_ARACHIDONIC_ACID_METABOLISM, KEGG_ALPHA_LINOLENIC_ACID_METABOLISM) are obviously enriched in the high-risk group, and three signaling cascades (KEGG_FRUCTOSE_AND_MANNOSE_METABOLISM, KEGG_PENTOSE_PHOSPHATE_PATHWAY KEGG_CITRATE_CYCLE_TCA_CYCLE) were abundant in the low-risk group by c2.cp.kegg.v7.2.symbols.gmt. These results suggest that this signature model may influence CRC progression and prognosis mainly through metabolism-related pathways