EPSILON-CP is a contact prediction server developed and maintained by the Robotics and Biology Laboratory at TU Berlin. EPSILON-CP uses deep neural networks to combine physicochemical (EPC-map), co-evolutionary and sequence-based information. EPSILON-CP achieves a mean precision of 36.1% for top-L/10 predicted long-range contacts on 21 hard CASP11 FM targets, which is 8% better than the current state-of-the-art algorithm MetaPSICOV. The improvement on 1.5L is 17%.