Table 3. Comparison of the performance of different methylation models in all samples of 9 kinds of cancers.

Accuracy	BLCAC	BLCAN	BRCAC	BRCAN	CESCC	CESCN	GBMC	GBMN	HNSCC	HNSCN
10 DMPs biomarker	0.022	0	0.05	0	0.05	0	0	0	0.02	0
4 CpGs of Article 1	0.593	0.095	0.584	0.01	0.663	0	0.327	0	0.742	0.02
5 CpGs of Article 2	0.939	0.333	0.92	0.388	0.977	0	0.928	0	0.977	0.977
12 CpGs of Article 3	0.872	0.19	0.749	0.143	0.625	0	0.242	0	0.87	0.02
Accuracy	LIHCC	LIHCN	LUADC	LUADN	LUSCC	LUSCN	UCECC	UCECN
10 DMPs biomarker	0.03	0	0.01	0	0	0	0.04	0
4 CpGs of Article 1	0.413	0	0.454	0	0.202	0	0.348	0
5 CpGs of Article 2	0.871	0.08	0.987	1	0.935	0.93	0.843	0.043
12 CpGs of Article 3	0.375	0	0.752	0	0.815	0	0.944	0.043

The vertical axis shows the different sources of the methylation models. The horizontal axis shows the tumor and normal methylation cohorts from 9 kinds of cancer. Affix C represents the tumor samples of the methylation data sets. Affix N represents the normal sample of the methylation cohorts. The numbers represent the percentage of samples predicted to be COAD in the total samples. The cohorts in the table are from 9 kinds of cancers, and the ideal result should be 0. Azuara D et al. [24] (Article 1)reported 4 CpG sites as diagnostic markers for COAD, and 4 of them had methylation values in the TCGA COAD cohort; Beggs AD et al. [25] (Article 2)report 6 CpG sites as diagnostic markers for COAD, and 5 of them had methylation values in the TCGA COAD cohort; Naumov VA et al. [26] (Article 3)reported 14 CpG sites as diagnostic markers for COAD, and 12 of them had methylation values in the TCGA COAD cohort.