Skip to main content
. Author manuscript; available in PMC: 2023 Dec 12.
Published in final edited form as: J Chem Inf Model. 2022 Nov 30;62(23):6069–6083. doi: 10.1021/acs.jcim.2c00879

Table 3:

Edge RBFEs obtained for the Cdk2 dataset calculated using ProFESSA. The table consists of average RBFEs from calculations with MCS, MCS-E, and MCS-Enw mapping algorithms. Median absolute deviations (MAD) are listed in parentheses.a

Edges ΔΔG Expt
U CCC 1 Expt 1 Expt iso
1h1q-1h1r −0.26 (0.08) −0.33 (0.04) 0.12 (0.05) −0.25 (0.02) 0.51
1h1s-1h1q 1.82 (0.00) 1.86 (0.00) 1.29 (0.01) 1.76 (0.09) 3.07
1h1s-1h1r 1.54 (0.22) 1.53 (0.03) 1.42 (0.07) 1.51 (0.07) 3.58
1h1s-1oiu −0.55 (0.15) −0.70 (0.09) 0.39 (0.01) 0.86 (0.09) 2.17
1oiu-1h1q 2.79 (0.02) 2.55 (0.08) 0.90 (0.00) 0.90 (0.00) 0.90
1oiu-1h1r 2.07 (0.12) 2.23 (0.08) 1.02 (0.05) 0.65 (0.02) 1.41
R2 0.01 0.02 0.50 0.74
MAE 1.56 1.57 1.08 1.03
RMSE 1.74 1.73 1.37 1.28
a

Listed are average relative free energy values using various network-wide analysis procedures: no cycle closure or experimental constraints (U); inclusion of cycle closure constraints (CCC); cycle closure constraints plus an additional experimental constraint for the most uncertain transformation identified through network Lagrange multiplier analysis (1 Expt); and further isolation of the uncertain ligand by removing all but one edge connection to the ligand (1 Expt iso). Summarized at the bottom are the linear correlation (R2), mean absolute error (MAE) and root-mean-square error (RMSE) with respect to experiment.