A dual-gene panel of two fragments of methylated IRF4 and one of ZEB2 in plasma cell-free DNA for gastric cancer detection

Figures & data

Figure 1. The design and workflow chart for identifying, selecting and validating the methylated cfDNA maker panel for gastric cancer.

Table 1. Patient and tumour characteristics of gastric adenocarcinoma in plasma validation.

	Plasma samples (a quarter sample volume)			Plasma samples (a normal sample volume)
	Gastric cancer	Gastric disease	Healthy	Gastric cancer	Control disease	Healthy
	(n = 37)	(n = 12)	(n = 43)	(n = 35)	(n = 39)	(n = 40)
Patient characteristics
Gender of patient
Male	27 (73%)	3 (25%)	29 (67%)	24 (69%)	15 (38%)	26 (65%)
Female	10 (27%)	9 (75%)	14 (33%)	11 (31%)	24 (62%)	14 (35%)
Age
Median	62	61	32	64	53	34
Range	(21–81)	(26–82)	(19–57)	(43–97)	(13–82)	(18–56)
Tumour characteristics		–	–		–	–
Tumour site
GE junction	2 (5%)			0 (0%)
Cardia/fundus/body	24 (65%)			23 (66%)
Antrum/pylorus	4 (11%)			5 (14%)
Diffuse	3 (8%)			3 (9%)
Not specified	4 (11%)			4 (11%)
Clinical stage^a		–	–		–	–
1	7 (19%)			2 (6%)
2	5 (14%)			6 (17%)
3	8 (22%)			8 (23%)
4	5 (14%)			3 (9%)
Undefined	12 (32%)			16 (46%)
Tumour size (cm)^b		–	–		–	–
≤4.0	16 (43%)			5 (14%)
4.1–7.9	7 (19%)			8 (23%)
8–11.9	1 (3%)			1 (3%)
≥12	0 (0%)			0 (0%)
Undefined	13 (35%)			21 (60%)
Histologic type^c		–	–		–	–
Intestinal	11 (31%)			10 (29%)
Diffuse	8 (22%)			3 (9%)
Mixed	8 (22%)			8 (23%)
Undefined	9 (25%)			14 (40%)
Disease type	–		–	–		–
Gastric polyps		6 (50%)
Nonatrophic		4 (33%)
Gastric mucosa with chronic inflammation		2 (17%)
Orthopaedic Disorders					36 (92%)
Kidney donor					3 (8%)

Note: ^a~Clinical stage (AJCC/UICC staging system).

^bLargest diameter.

^cBased on Lauren classification.

Figure 2. DNA methylation levels (β values) of the five DMCs in healthy WBC, normal gastric tissues from GEO datasets and gastric cancer tissues from TCGA dataset of STAD. The abscissa is the type of sample, and the ordinate is the methylation level (β value). The extremes of the boxes define the upper and lower quartiles, and the centre lines define the median. Statistical analysis between the two groups was performed using Wilcoxon rank sum test, with p < 0.05 indicating significance.

Figure 3. The methylation levels of five markers were analysed in the white blood cells of healthy individuals and gastric cancer tissues using 2^−ΔΔCt values or ct values. The sample size for WBC was n = 36, and for GC it was n = 40. The upper picture displays the methylation levels based on 2^−ΔΔCt values, while the lower picture shows the methylation levels based on ct values. Statistical analysis between the two groups was performed using Mann-Whitney test, with p < 0.05 indicating significance.

Table 2. The detection rates of markers in healthy white blood cells and gastric cancer tissues.

Marker	Healthy white blood cells (n = 36)		GC (n = 40)
	Positive	Negative	Positive	Negative
IRF4–1	2 (5.6%)	34 (94.4%)	40 (100.0%)	0 (0.0%)
IRF4–2	2 (5.6%)	34 (94.4%)	40 (100.0%)	0 (0.0%)
ZEB2	1 (2.8%)	35 (97.2%)	36 (90.0%)	4 (10.0%)
ITGA4	0 (0.0%)	36 (100.0%)	39 (97.5%)	1 (2.5%)
DNAJC6	19 (52.8%)	17 (47.2%)	40 (100.0%)	0 (0.0%)

Figure 4. The methylation levels of selected markers in a quarter sample volume of plasma were analysed using 2^−ΔΔCt values or Ct values. The study included healthy individuals, those with benign gastric disease, and those with GC. The upper picture displays the methylation levels based on 2^−ΔΔCt values, while the lower picture shows the methylation levels based on Ct values. Statistical analysis between two groups was performed using Mann Whitney-test, and p < 0.05 was considered significant.

Table 3. The diagnostic performance of markers was evaluated using a quarter sample volume of plasma samples analysed based on 2^−∆∆Ct values or Ct values.

Indicator	Marker/marker combination	AUC	Youden index	AUC 95% CI	Sen^b	Spec	Method for combination
2^−∆∆Ct	IRF4–1	0.587	0.377	0.459–0.695	43.2	94.5	/
	ITGA4	0.470	0.269	0.333–0.633	32.4	94.5
	IRF4–2	0.453	0.377	0.314–0.599	43.2	94.5
	ZEB2	0.396	0.207	0.274–0.539	24.3	96.4
	IRF4–1+ZEB2+ITGA4	0.686	0.377	0.573–0.775	43.2	94.5	Logistic regression
	IRF4–1+ZEB2^a	0.670	0.377	0.563–0.797	43.2	94.5
	IRF4–1+ITGA4	0.621	0.377	0.492–0.727	43.2	94.5
	IRF4–1+IRF4–2+ZEB2	0.608	0.216	0.547–0.662	21.6	100.0
	IRF4–1+IRF4–2+ZEB2+ITGA4	0.608	0.216	0.554–0.676	21.6	100.0
	IRF4–2+ZEB2+ITGA4	0.608	0.216	0.547–0.662	21.6	100.0
	IRF4–2+ZEB2	0.581	0.162	0.514–0.635	16.2	100.0
	IRF4–1+IRF4–2+ITGA4	0.567	0.450	0.401–0.715	54.1	90.9
	IRF4–1+IRF4–2	0.537	0.413	0.389–0.720	48.6	92.7
	ZEB2+ITGA4	0.492	0.269	0.363–0.624	32.4	94.5
	IRF4–2+ITGA4	0.463	0.377	0.329–0.608	43.2	94.5
Ct	IRF4–2	0.711	0.404	0.617–0.779	45.9	94.5	/
	IRF4–1	0.689	0.377	0.612–0.769	43.2	94.5
	ITGA4	0.639	0.269	0.571–0.719	32.4	94.5
	ZEB2	0.588	0.207	0.517–0.679	24.3	96.4
	IRF4–1+IRF4–2+ITGA4	0.779	0.522	0.709–0.866	64.9	87.3	Logistic regression
	IRF4–1+IRF4–2	0.759	0.486	0.653–0.850	59.5	89.1
	IRF4–1+IRF4–2+ZEB2+ITGA4	0.757	0.495	0.641–0.850	62.2	87.3
	IRF4–1+IRF4–2+ZEB2	0.740	0.477	0.647–0.830	56.8	90.9
	IRF4–2+ZEB2+ITGA4	0.726	0.441	0.629–0.814	56.8	87.3
	IRF4–2+ITGA4	0.719	0.404	0.623–0.810	51.3	89.1
	IRF4–2+ZEB2	0.712	0.422	0.626–0.788	51.3	90.9
	IRF4–1+ITGA4	0.708	0.395	0.630–0.790	48.6	90.9
	IRF4–1+ZEB2+ITGA4	0.701	0.395	0.619–0.800	48.6	90.9
	IRF4–1+ZEB2	0.694	0.395	0.594–0.780	48.6	90.9
	ZEB2+ITGA4	0.661	0.332	0.586–0.758	40.5	92.7
Ct	IRF4–1+IRF4–2+ZEB2+ITGA4	/	0.530	/	67.6	85.4	1/n algorithm
	IRF4–1+IRF4–2+ITGA4		0.521		64.9	87.3
	IRF4–1+IRF4–2		0.504		59.5	90.9
	IRF4–1+IRF4–2+ZEB2		0.494		62.2	87.3
	IRF4–2+ZEB2+ITGA4		0.440		56.8	87.3
	IRF4–1+ZEB2+ITGA4		0.431		54.0	89.1
	IRF4–2+ZEB2		0.423		51.3	90.9
	IRF4–2+ITGA4		0.404		51.3	89.1
	IRF4–1+ZEB2		0.396		48.6	90.9
	IRF4–1+ITGA4		0.396		48.6	90.9
	ZEB2+ITGA4		0.333		40.5	92.7

Note:^aThe sensitivity and specificity corresponding to the second maximum Youden index due to its high sensitivity.

^bSen is sensitivity, and spec is specificity.

Figure 5. The methylation levels of selected markers in plasma samples and ROC curves. (a) displays the methylation levels based on 2^−ΔΔCt values, while (b) displays the methylation levels based on ct values. CD represents control disease and GC represents gastric cancer. Statistical analysis between two groups was performed using Mann-Whitney test, and p < 0.05 was considered significant. (c) are ROC curves of single marker and marker combination based on 2^−ΔΔCt values (left) and ct values (right). Combination represents IRF4-(1&2) and ZEB2.

Table 4. The diagnostic performance of markers in plasma validation analysed based on 2^−∆∆Ct or ct values.

Indicator	Marker/marker combination	AUC (95% CI)	Sen^b	Spec	Cutoff	Method for combination
2^−∆∆Ct	IRF4(1&2)	0.865 (0.797–0.956)	77.1	93.7	0.0000070	/
	ZEB2	0.731 (0.630–0.831)	51.4	94.9	0.0000095	/
	IRF4(1&2) and ZEB2	0.857 (0.740–0.946)	77.1	92.4	0.1491	Logistic regression
Ct	IRF4(1&2)	0.841 (0.751–0.907)	71.4	93.7	45.8	/
	ZEB2	0.734 (0.621–0.847)	45.7	98.7	42.3	/
	IRF4(1&2) and ZEB2^a	0.850 (0.772–0.918)	74.3	92.4	0.3922	Logistic regression
	IRF4(1&2) and ZEB2	/	74.3	92.4	45.8, 42.3	1/n algorithm

Note:^aThe sensitivity and specificity corresponding to the second maximum Youden index due to its high sensitivity.

^bSen is sensitivity, and spec is specificity.

Data availability statement

The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.