Fig. 5: Biomarker identification and validation using ATAC-seq in the BCPAP cells treated with 5-Aza and machine learning. | npj Precision Oncology

Fig. 5: Biomarker identification and validation using ATAC-seq in the BCPAP cells treated with 5-Aza and machine learning.

From: Epigenomic profiling of papillary thyroid carcinoma reveals distinct subtypes with clinical implications

Fig. 5

A Heatmap displaying the differential accessibility of regions upon 5-Aza treatment. Open DAR means more accessible regions upon BCPAP treatment with 5-Aza. B Six key biomarkers were selected through machine learning of TCGA THCA RNA-seq data to distinguish between normal tissue and PTC subtypes. C ROC curves evaluating the performance of the random forest model in classifying DNA methylation subgroups and normal tissue. The balanced accuracy scores, computed as the average of sensitivity and specificity, were 0.89 for normal tissue, 0.86 for PTC1, and 0.88 for PTC2. D Integrative multi-omic status at the AGAP2 locus.

Back to article page