We obtained a total of 2,836 differential methylation CpGs (DMCs) in CRC tumor tissues compared with adjacent normal tissues, including 2,113 hyper-methylation and 723 hypo-methylation ones. Figure S1A illustrated the differential methylation landscape of all CpGs, and Figure S1B showed the methylation level of the 2,836 DMCs in CRC tumor and adjacent normal tissue samples as a heatmap.
Crc Prognostic Model
We used the univariate Cox regression analysis on 287 CRC patients and identified 53 DMCs significantly associated with OS. The coefficients of those 53 CpGs was illustrated in Figure S2A. LASSO-Cox was further established and determined 16 optimal prognostic genes, including cg01129320, cg01992382, cg03904639, cg05317090, cg06084210, cg09170112, cg09492451, cg09918510, cg11712188, cg12740527, cg14675211, cg15265085, cg16834823, cg18237607, cg18624636 and cg19611175. Besides, the Lambda value to turn parameter in the LASSO-Cox was in Figure S2B. And the risk score was as follow:
Risk Score = 1.1175 × β Value of cg01129320 + 0.049 × β Value of cg01992382 + 0.2382 × βValue of cg03904639 + 0.4814 × βValue of cg05317090 + 0.8733 × βValue of cg06084210 + 0.5062 × βValue of cg09170112 + 0.4175 × βValue of cg09492451 + 0.0327 × βValue of cg09918510 + 1.6943 × βValue of cg11712188 + 1.4206 × βValue of cg12740527 + 2.9668 × βValue of cg14675211 + 0.4418 × βValue of cg15265085 + 0.0137 × βValue of cg16834823 + 0.557 × βValue of cg18237607 + 0.4412 × βValue of cg18624636 + 0.9536 × βValue of cg19611175.
Risk score was a good assessment of patient survival prognosis
Patients were divided into high and low risk groups based on cut-off = 6.01. Kaplan-Meier curve showed that high-risk group had significantly longer survival compared to low-risk groups. Besides, we found that the AUCs of the 1-, 3-, and 5-year OS in the training set were 0.895, 0.89, and 0.959, respectively; the AUCs of the 1-, 3-, and 5-year OS in the first training set were 0.625, 0.659, and 0.74; the AUCs of OS at 1 year, 3 years, and 5 years in the second training set were 0.615, 0.565, and 0.65 respectively; the AUCs of OS at 1 year, 3 years, and 5 years in the whole training set were 0.709, 0.697, and 0.793, indicating that risk score could better predict patients 1-, 3-, and 5-year survival rates (Fig. 1A-1D). Collectively, the above results indicated that the established prognostic model according to these 16 CpG sites performed well in terms of survival prognosis.
Risk Score Was Independent Of Other Prognostic Factors
The 300 samples were grouped according to age, gender, and stage and their risk scores were calculated. The results revealed that there was a significant difference in risk scores between different ages, however, no significant difference could be obtained in genders and stages (Fig. 2A-2C). Next, multivariate Cox regression was established and proved the strong independence of our prognostic model including age, gender, and stage factors through the survival package in R (Fig. 2D). In addition to the risk score, stage was also an independent prognostic factor.
Nomogram could better predict the 1-year 3-year 5-year survival of patients.
The nomogram constructed using two independent prognostic factors, stage and risk score to predict the 1-, 3-, and 5-year OS of CRC patients was in Fig. 3A. Besides, the calibration chart to assess the accuracy of the nomogram was displayed in Fig. 3B, which proved the nomogram might infer estimates of slightly higher or lower actual survival probability. Moreover, the 1-year, 3-year, 5-year AUC of the combined model was higher compared with a single factor, indicating that a nomogram established with all independent prognostic factors could better predict the patient's 1-year 3-year 5-year survival than risk score and stage (Fig. 4).