Algorithm 1 Rotate through cell types for training
Require: C (set of training cell types)
Require: R (genomic regions selected for training)
1:	for i in R do
2:	for k in C do
3:	xi ← xij ∀j ∈ C, j (xij are cell type specific features as
	defined in Equation 4)
4:	yik ← ChIP-seq peaks for cell type k in region i
5:	loss = calculateLoss(xi, yik)
6:	updateGradients(loss)
7:	if model has converged then
8:	exit
9:	end if
10:	end for
11:	end for