Fig. 2
From: Biologically relevant transfer learning improves transcription factor binding prediction
Heatmap of pairwise cosine similarities between the binding vectors of 162 TFs (SMAD3 is excluded because it had no bound regions). Homologous TFs from the same families form well-defined clusters along the heatmap’s diagonal. Highlighted with white arrows, an example of cooperativity for highly correlated TFs from different families, namely GATA1, GATA2, SOX6, and TAL1, which are required for erythropoiesis [40]. TEAD4 is also highly correlated with these four TFs (teal arrows), in accordance with previous literature [41]