Fig. 4: Cancer detection by multimodal analysis of cfDNA WMS data. | Nature Communications

Fig. 4: Cancer detection by multimodal analysis of cfDNA WMS data.

From: Multimodal analysis of cell-free DNA whole-methylome sequencing for cancer detection and localization

Fig. 4

a Visualization of individual modalities by tSNE plots (n = 1277). Healthy control individuals and patient samples from different cancer types are annotated by shapes and colors. b Receiver operator characteristics for the classification of cancer patients (training n = 542; test n = 238) from healthy controls (training n = 352; test n = 145) by individual modalities and the integrative THEMIS model for the training and the test cohort respectively. The 95% confidence intervals (CIs) for AUCs are calculated with 2000 stratified bootstrap replicates. c Venn diagrams depicting the number of healthy individuals misclassified as cancer patients (false positives) and the number of cancer patients correctly identified (true positives) at a specificity of 99% by individual modalities and THEMIS. Samples from both the training and the test cohort are included. Source data are provided as a Source Data file.

Back to article page