Fig. 2: Validation of hallmark annotations in synthetic training data. | Communications Biology

Fig. 2: Validation of hallmark annotations in synthetic training data.

From: OncoMark: a high-throughput neural multi-task learning framework for comprehensive cancer hallmark quantification

Fig. 2

Heatmaps depict the log2fold change and −log10FDR values for hallmark gene sets, comparing hallmark-present versus hallmark-absent samples for each hallmark. Deep blue intensity in both the log2fold change and −log10FDR panels indicates strong and significant enrichment of the gene set in hallmark-positive samples, supporting accurate hallmark annotation. Blank areas in the −log10FDR heatmap indicate infinite FDR values (i.e., unadjusted p values of zero), corresponding to highly significant associations.

Back to article page