Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: Dualmarker: a flexible toolset for exploratory analysis of combinatorial dual biomarkers for clinical efficacy

Fig. 4

De-novo identification of marker2 from gene expression to combine with marker1 (mut_ARID1A). The dual marker Cox regression model was compared with the single marker model using marker1 and marker2 separately. The -log10 p values of the LRT test between models are shown on the x-axis and y-axis, and the dashed line indicates the p value = 0.01 (a). Top significant marker2s to combine with mut_ARID1A according to the significance of LRT in model comparison of the dual-marker model versus the mut_ARID1A single-marker model; dashed lines show p value = 0.01 and 0.05(b). The signed -log10-p value is provided, and the sign is the direction of the effect of marker2 on OS as a single marker. The arrow points to the known biomarker partners, CXCL13, IFNG and the novel gene HMGB1, which showed a statistical interaction with mut_ARID1A (c-e). Four quadrants/groups divided by mut_ARID1A and HMGB1 expression levels (high vs low, cut by population median) had significantly different OS values (c, p value = 0.0019, log-rank test for all group comparisons). ARID1A-mutated patients had longer OS under the condition of high HMGB1 but not low HMGB1 (d). Median survival time and confidence interval for each quadrant/group (e)

Back to article page