Fig. 2
From: ADAGE signature analysis: differential expression analysis with data-defined gene sets
ADAGE model and gene signatures. In an ADAGE model, every gene is linked to every node through an edge. The edge weight is fixed after model training and its magnitude is reflected by edge thickness. The distribution of gene weights to a node is centered at zero and close to normal. Genes giving weights higher than the positive high-weight (HW) cutoff together form the positive gene signature for that node (genes in the orange circle). Similarly, genes giving weights lower than the negative HW cutoff together form negative gene signature for that node (genes in the blue circle)