PaperCombining docking pose rank and structure with deep learning improves protein−ligand binding mode prediction over a baseline docking approach