Skip to main content

Table 1 Cross-dataset classification performances.

From: Use B-factor related features for accurate classification between protein binding interfaces and crystal packing contacts

Training dataset Feature Tested datasets
   BNCPCS DC Bahadur Ponstingl
BNCPCS ΣB 0.93(0.97) 0.32(0.65) 0.65(0.82) 0.82(0.91)
  ΔASA 0.92(0.96) -0.18(0.47) 0.59(0.78) 0.73(0.86)
  avgΣB 0.92(0.96) 0.37(0.68) 0.64(0.82) 0.80(0.90)
  avgNo.B 0.95(0.98) 0.25(0.60) 0.70(0.84) 0.84(0.92)
  avgΣB*avgNo.B 0.94(0.97) 0.33(0.66) 0.70(0.85) 0.82(0.91)
  avgΔASA 0.91(0.96) -0.16(0.48) 0.64(0.81) 0.72(0.86)
DC ΣB 0.85(0.92) 0.38(0.69) 0.68(0.85) 0.81(0.90)
  ΔASA 0.73(0.86) 0.15(0.57) 0.66(0.84) 0.62(0.80)
  avgΣB 0.88(0.94) 0.45(0.73) 0.73(0.87) 0.80(0.90)
  avgNo.B 0.80(0.90) 0.46(0.72) 0.74(0.87) 0.70(0.84)
  avgΣB*avgNo.B 0.86(0.93) 0.45(0.73) 0.75(0.88) 0.81(0.90)
  avgΔASA 0.76(0.88) 0.27(0.63) 0.68(0.85) 0.66(0.82)
Bahadur ΣB 0.84(0.92) 0.38(0.69) 0.71(0.86) 0.79(0.89)
  ΔASA 0.73(0.86) 0.15(0.57) 0.66(0.84) 0.62(0.80)
  avgΣB 0.84(0.92) 0.41(0.70) 0.75(0.88) 0.81(0.90)
  avgNo.B 0.86(0.93) 0.33(0.66) 0.75(0.88) 0.77(0.88)
  avgΣB*avgNo.B 0.88(0.94) 0.45(0.73) 0.77(0.89) 0.83(0.91)
  avgΔASA 0.81(0.90) 0.21(0.60) 0.69(0.85) 0.69(0.84)
Ponstingl ΣB 0.88(0.94) 0.39(0.70) 0.69(0.85) 0.81(0.90)
  ΔASA 0.91(0.96) -0.18(0.47) 0.59(0.79) 0.72(0.86)
  avgΣB 0.90(0.95) 0.43(0.71) 0.73(0.87) 0.82(0.91)
  avgNo.B 0.95(0.98) 0.25(0.60) 0.70(0.84) 0.84(0.92)
  avgΣB*avgNo.B 0.90(0.95) 0.40(0.70) 0.75(0.88) 0.83(0.92)
  avgΔASA 0.92(0.96) -0.19(0.46) 0.65(0.82) 0.78(0.89)
PDBbind ΣB 0.93(0.97) 0.38(0.68) 0.62(0.79) 0.72(0.86)
  ΔASA 0.88(0.94) -0.16(0.48) 0.49(0.68) 0.62(0.79)
  avgΣB 0.88(0.94) 0.41(0.71) 0.71(0.86) 0.83(0.92)
  avgNo.B 0.92(0.96) 0.38(0.68) 0.74(0.88) 0.80(0.90)
  avgΣB*avgNo.B 0.90(0.95) 0.38(0.69) 0.76(0.88) 0.86(0.93)
  avgΔASA 0.88(0.94) 0.02(0.51) 0.66(0.84) 0.70(0.85)
  1. X.XX(Y.YY) represent the classification performances where X.XX is the MCC score and Y.YY is the accuracy score. The italic numbers are the learning performances, and thus they are not used in the comparison. The bold-font numbers are the better performances when comparing ΣB and avgΣB*avgNo.B with ΔASA, and ΣB with ΔASA.