Skip to main content

Table 2 Prediction accuracy and cardinality for the best ten models obtained by Soto’s method [5]

From: Visual analytics in cheminformatics: user-supervised descriptor selection for QSAR methods

Model Predictive accuracy Cardinality
M1 (Mn/MW, Sp, RHyDp, ETA_EtaP_F_L) R2 = 0.26
MAE = 4.62
RMSE = 8.14
4
M2 (Mn/MW, MDEO-11, D/Dr09, SMTIV) R2 = 0.32
MAE = 5.94
RMSE = 8.31
4
M3 (Mn/MW, nHBint4, nHBint10, ETA_dEpsilon_B) R2 = 0.56
MAE = 4.03
RMSE = 6.22
4
M4 (Mn/MW, nsCH3, nF6Ring, ALOGP2, RDCHI) R2 = 0.41
MAE = 3.94
RMSE = 6.75
5
M5 (Mn/MW, nROH, n6Ring, nHCsatu, ALOGP2) R2 = 0.68
MAE = 3.28
RMSE = 5.78
5
M6 (Mn/MW,nP, minHBa, T(O..P), ETA_Epsilon_3) R2 = 0.25
MAE = 4.48
RMSE = 7.20
5
M7 (Mn/MW, ETA_dEpsilon_B, C-005, SHaaCH, nHBint9,nCt) R2 = 0.31
MAE = 4.19
RMSE = 7.20
6
M8 (Mn/MW, ndssC, minHBint9, MSD, C-004, Mw/Mn (PDI), crosshead speed(CHS)) R2 = 0.39
MAE = 3.92
RMSE = 6.86
7
M9 (Mn/MW, Pol, Wap, maxHAvin, nHAvin, MWC04) R2 = 0.15
MAE = 4.92
RMSE = 7.88
6
M10 (Mn/MW,maxHBint6, ETA_dEpsilon_A, TIC2, ndO, nHdCH2) R2 = 0.48
MAE = 4.02
RMSE = 7.09
6
  1. The second column shows the predictive accuracy of the “best” model after applying 4-fold cross validation on three different methods (linear regression, decision trees, and neural networks). The parameter setup and predictive accuracy for all methods is available in the Additional file 1: Table S2.