Skip to main content
. 2022 Mar 22;7(13):11057–11067. doi: 10.1021/acsomega.1c07037

Table 3. Summary of Cross-Validation Results of the Combinations of Three Energy Models and the GB Model Used for Energy Minimizationa.

    training
 testing
GBb EMc PCC RMSE PCC RMSE
OBC-II model 1d 0.440 ± 0.045 3.065 ± 0.104 0.496 ± 0.241 3.245 ± 0.672
  model 2e 0.513 ± 0.044 2.968 ± 0.094 0.519 ± 0.178 3.295 ± 0.832
  model 3f 0.596 ± 0.040 2.818 ± 0.098 0.651 ± 0.151 2.958 ± 0.894
GBneck model 1 0.442 ± 0.045 3.061 ± 0.106 0.495 ± 0.242 3.246 ± 0.678
  model 2 0.520 ± 0.044 2.955 ± 0.098 0.522 ± 0.182 3.313 ± 0.867
  model 3 0.653 ± 0.042 2.657 ± 0.151 0.609 ± 0.205 3.175 ± 1.130
GBneck2 model 1 0.447 ± 0.045 3.053 ± 0.106 0.503 ± 0.241 3.255 ± 0.694
  model 2 0.527 ± 0.044 2.941 ± 0.010 0.529 ± 0.180 3.306 ± 0.884
  model 3 0.675 ± 0.040 2.586 ± 0.135 0.653 ± 0.151 3.091 ± 1.103
a

Summary of PCC and RMSE over training (randomly selected 80%) and testing (remaining 20%) of the PPI-46 data set for the three energy models (eqs 1–3) and the GB model used in energy minimizations with ϵin = 1, repeated 25 times; mean and standard deviation are provided.

b

GB model.

c

Energy model.

d

MM/PB model (eq 1).

e

MM/PBSA model (eq 2).

f

MM/PBSA/E model (eq 3) with parameters σ = 1.20, r = 0.8, and cutoff radius 4.0 Å.