Skip to main content
Fig. 5 | BMC Bioinformatics

Fig. 5

From: DeepGRN: prediction of transcription factor binding site across cell-types using attention-based deep neural networks

Fig. 5

Analysis of attention weights and saliency scores. (a) Scatterplot of attention weights from positive strand and reverse strand. (b) Scatterplot of saliency scores from positive strand and reverse strand. (c) Scatterplot of ChIP-Seq fold change and mean attention weights from both strands. Z-score transformation is applied to both axes. (d) Distribution of the correlation between attention weights/saliency scores and ChIP-Seq fold change. The dashed line represents the mean of each group. The p-value is calculated using the Wilcoxon signed-rank test. The attention weights and saliency scores on the reverse complement strand are reversed before plotting. ρ: Spearman Correlation Coefficient, σ: Pearson Correlation Coefficient. The correlation between normalized ChIP-Seq Fold change and normalized saliency scores is 0.40 (Spearman) and 0.49 (Pearson)

Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us