Table 1 Methods for defining seven pairs of positive/negative training sets in a cell type with STARR-seq data available
Methods | Labels | Size (sequences) | Positive set | Negative set |
|---|---|---|---|---|
CRM+TF+/Non-CRM | TF binding | 17,558~272,128 | CRMs overlapping TF binding peaks | Randomly selected non-CRMs |
CRM+TF+/CRM+S− | TF binding | 17,558~272,128 | CRMs overlapping TF binding peaks | CRMs that cannot be predicted to be active and do not overlap STARR peaks |
CRM+S+/Non-CRM | STARR peaks | 22,610~71,176 | CRMs overlapping STARR peaks | Randomly selected non-CRMs |
CRM+S+/CRM+S− | STARR peaks | 22,610~71,176 | CRMs overlapping STARR peaks | CRMs that cannot be predicted to be active and do not overlap STARR peaks |
Bin+S+/Bin+S− | STARR peaks | 60,668~109,118 | 700bp bin overlapping STARR peaks | 700bp bin not overlapping STARR peaks |
Bin+ac+/Bin+ac− | H3K27ac peaks | 175,530~1,688,868 | 700bp bin overlapping H3K27ac | 700bp bin not overlapping H3K27ac |
Bin+S+&ac+/Bin+S−&ac− | STARR & H3K27ac peaks | 7220~49,462 | 700bp bin overlapping STARR&H3K27ac peaks | 700bp bin not overlapping STARR&H3K27ac peaks |