Neph et al describes a procedure which involves locating putative TF binding sites in DNaseI-seq footprinted regions. There's a paper by Maurano et al that shows footprint signal have good concordance with ChIP-seq signal. You then do bedops set operations to find overlaps with gene promoter regions, i.e. where a TF can bind to a gene's promoter may indicate a regulatory relationship between the TF and the target gene.