Fig. 8From: GODoc: high-throughput protein function prediction using novel k-nearest-neighbor and voting algorithmsAn illustration showing 1NN, Fixed-KNN, and Dynamic-KNN. A query protein is colored in gray. There are two GOs for training data (colored in green and orange) where the neighbor proteins picked are indicated with a solid circle. a 1NN: the nearest training protein is selected. b Fixed-KNN: three training proteins are picked for Kâ=â3. c Dynamic-KNN: five proteins are selected, as their distances to the query are smaller than the threshold dBack to article page