Abstract: With the gap between the sequence data and their functional annotations becomes increasing wider, many computational methods have been proposed to annotate functions for unknown proteins.