jps.21745

[ Pobierz całość w formacie PDF ]

�sterberg249 PLS 4 descriptors, 3 hydrogen bonding related and log P 35 34 0.71 0.68a 0.37q
�sterberg249 PLS 4 descriptors, 3 hydrogen bonding related and log P 23 22 0.72 0.68a 0.50q
Norinder250 PLS Eletrotopological indices (E-state descriptors) 58 NAb 0.80 0.77a NAb
Norinder250 PLS Eletrotopological indices (E-state descriptors) 28 30 0.75 0.70a 0.37q
Norinder246,250 PLS MolSurf descriptors 28 30 0.86 0.78a 0.35q
Young208 MLR 1 descriptor, Dlog P as (log Poct log Pcycl)20 NAb 0.69 NAb NAb
Kaliszan240 MLR 2 descriptors, MW and log Pcycl 20 NAb 0.85 NAb NAb
Platts241 MLR 6 descriptors, LFER &Iir 148 NAb 0.75 0.71a NAb
Platts241 MLR 6 descriptors, LFER & Iir 74 74s 0.76 NAb 0.72s
(Continued)
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 12, DECEMBER 2009
IN VIVO, IN VITRO, AND IN SILICO METHODS
4449
Table 3. (Continued )
Quantitative Study: Training Test
log BB Estimation Method Number Descriptors Set Set r2 q2 pred-r2
Platts241 MLR 6 descriptors, LFER & Iir 118 30t 0.75 0.71a 0.73t
Luco247 PLS 18 descriptors 58 14 0.85 0.75v 0.24q
Luco247 PLS 18 descriptors 58 25 0.85 0.75v 0.54q
Luco247 PLS 18 descriptors 58 27w 0.85 0.75v 92.3j
Clark211 MLR 2 descriptors, Clog P and PSA 55 7 0.79 NAb 0.40x
Clark211 MLR 2 descriptors, Clog P and PSA 55 5 0.79 NAb 0.23x
Liu.242 MLR 2 descriptors, MW and LA (molecular lipoaffinity) 55 11 0.79 0.76a 0.84
Liu242 Backpropagation Neural 3 descriptors (TPSA, LA, and MW) 55 11 0.81 NAb 0.81
Network (ANN)
Hou315 Genetic Algorithm & MLR 4 descriptors, out of 27, identified by this method 59 12 0.76 0.71a 0.88
Hou315 Genetic Algorithm & MLR 4 descriptors, out of 27, identified by this method 59 23 0.76 0.71a 0.64
Subramanian255 Genetic Algorithm & PLS 7 descriptors 58 39 0.85 0.82a 0.61
Subramanian255 Genetic Algorithm & PLS 7 descriptors 58 181w 0.85 0.82a 66.2j
Subramanian255 Genetic Algorithm & PLS 7 descriptors 58 181y 0.85 0.82a 64.5j
Subramanian255 Genetic Algorithm & PLS 7 descriptors 58 181p 0.85 0.82a 63.3j
Stanton251 PLS 4 descriptors (including one of new HSA), out of 161 97 NAb 0.78 0.75a NAb
Winkler238 Bayesian Neural Network 7 descriptors, out of multiple property-based 85 21 0.81 NAb 0.65
descriptors, identified by ARD
Cabrera243 MLR 3 TOPS-MODE descriptors 114 NAb 0.70 0.66a NAb
Cabrera243 MLR 3 TOPS-MODE descriptors 86 28k 0.33z NAb 0.33l
Cabrera243 MLR 3 TOPS-MODE descriptors 114 28i 0.70 0.66a 85.7j
Yap262 Bayesian Neural Network 7 descriptors, out of 1497 descriptorsab 129 30 NAb NAb 0.70
Garg224 Backpropagation Neural 18 descriptors 132 50 0.81 NAb 0.79
Network (ANN)
Abraham189 MLR 7 descriptors. LFER and Ic and Ivm 328 NAb 0.75 NAb NAb
Abraham189 MLR 7 descriptors. LFER and Ic and Ivm 164 164 0.71 NAb 0.25u
Obrezanova266 Gaussian Processac 7 descriptorsad 85 21 0.61 NAb 0.74
Obrezanova266 Gaussian Processae 7 descriptorsad 85 21 0.69 NAb 0.81
Iyer236 Genetic Algorithm & MLR 5 descriptors, out of multiple molecular 56 7 0.85 0.80a 0.68
descriptorsaf
Iyer236 Genetic Algorithm & MLR 5 descriptors, out of multiple molecular 46 10ag NAb NAb 0.70ah
descriptorsaf
Pan237 Genetic Algorithm & MLR 2 descriptors (Clog P & PSA), out of multiple 150 NAb 0.69 0.60a NAb
molecular descriptorsaf
Pan237 Genetic Algorithm & MLR 2 descriptors (Clog P & PSA), out of multiple 104 46 0.69 0.64a 0.53ai
molecular descriptorsaf
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 12, DECEMBER 2009
4450
MENSCH ET AL.
DOI 10.1002/jps
Table 3. (Continued )
Quantitative Study: Training Test
log BB Estimation Method Number Descriptors Set Set r2 q2 pred-r2
Pan237 Genetic Algorithm & MLR 2 descriptors, out of multiple molecular 24aj 13aj 0.85 0.83a 0.76ak
descriptorsaf
Pan237 Genetic Algorithm & MLR 4 descriptors, out of multiple molecular descriptorsaf 63aj 25aj 0.69 0.66a 0.79ak
Pan237 Genetic Algorithm & MLR 3 descriptors, out of multiple molecular descriptorsaf 17aj 8aj 0.80 0.72a 0.92ak
Labute234 PCR 15 descriptors, out of 32 VSA 75 NAb 0.83 0.73a NAb
Sun252 PLS Atom types as molecular descriptors 57 13 0.91 0.50al 0.33am
Abraham195 h MLR 5 descriptors. LFER 30 NAb 0.87 NAb NAb
a
Internal validation by leave-one-out (LOO) cross-validation.
b
NA, no data available.
c
AS fuse information of both the isomorphic similarity and nonisomorphic dissimilarity.
d
5 random independent test sets were extracted from initial set of 130 compounds.
e
Considering the prediction for all the compounds which have participated once in test stages (3 22 � 2 21) not involved in their corresponding training sets.
f
Three first descriptors identify by other studies as key.216 Ic and Iv linked to the nature of the data set.174
g
Pred-r2 was not reported, but the root-mean-squared error of prediction: RMSEP. Monte Carlo Crossvalidation Method, leave-group-out approach; with N � 30 000 and nv � n/
218 was used.
h
log PS estimation.
i
Qualitative Test Set: BBB�/BBB . Using BBBpred as the discriminant variable, a threshold value of 0.00 was fixed to delineate BBB� and BBB compounds.
j
Test Set Overall Accuracy.
k
Five random independent test sets were extracted from initial set.
l
Pred-r2 was not reported, but the overall mean absolute error for the five groups.
m
Ic and Iv linked to the nature of the data set.174
n
Rational selection. Physchem properties that may be involved in passive diffusion: log P, PSA, and MW.
o
Rational selection. Physchem properties that may involved in passive diffusion including amphipilic components of the SASA, determined from MC simulations.
p
Qualitative Test Set: BBB�/BBB . Using BBBpred as the discriminant variable, a threshold value of 0.4 was fixed to delineate BBB� and BBB compounds.
q
Pred-r2 was not reported, but the root-mean-square error for the dependent variable: RMSE.
r
Ii an indicator variable that is set to 1 for a compound containing a carboxylic acid fragment and 0 otherwise.
s [ Pobierz całość w formacie PDF ]

download ^ pobieranie ^ pdf ^ do ÂściÂągnięcia ^ ebook

Pokrewne