Fig. 5
From: Integrative prediction of gene expression with chromatin accessibility and conformation data
a Model performance assessed in terms of Spearman correlation on hold-out test data for models including TF-gene scores computed in the promoter and in the distal enhancers. Generally, including TF predictions improves model performance compared to considering only peak features. Significance is assessed using a Wilcoxon test using the promoter model as the reference group (****\(p <0.0001\), **\(p <0.01\), *\(p <0.05\), ns : \(p \ge 0.05\)). b UpSet plot showing the relationship between TFs with a non-zero regression coefficients inferred by the extended feature space models. c Ranking of the top 20 TFs by their absolute regression coefficients for each cell line. The colour code indicates the mean scaled regression-coefficient of the TFs computed in a ten-fold outer cross-validation