Skip to main content

Table 2 The best parameter combination trained from CAFA2-Swiss is utilized in the CAFA2-Benchmark

From: GODoc: high-throughput protein function prediction using novel k-nearest-neighbor and voting algorithms

Type PCA dims Exp. ratio (%) Method Voting weight k or dynamic
BPO 107 96.0 1NN 1
Fixed-KNN inverse 7
Dynamic-KNN inverse Q2
Hybrid-KNN inverse 4 + Q2
CCO 51 95.0 1NN 1
Fixed-KNN inverse 9
Dynamic-KNN inverse Q2
Hybrid-KNN inverse 7 + Q2
MFO 121 96.5 1NN 1
Fixed-KNN inverse 3
Dynamic-KNN inverse Q2
Hybrid-KNN inverse 3 + Q2