This slide is specially use For Pharmacy student

1. QSPR Approach to Drug
Disposition Prediction
1
Miss. Shatavari Bhosale
M. Pharm 1st Year
(Pharmaceutics)
BHARATI VIDYAPEETH
College of Pharmacy, Kolhapur

2.  Introduction
 Why QSPR It is Useful
 Method in QSPR Model
 QSPR Approaches
 Current Method of QSPR
 Software
 Conclusion
 Reference
2
Contents

3. Introduction
3
 QSPRs are computer-assisted mathematical models, which relate t
he physico-chemical property of compounds to their chemical stru
cture QSPR studies are performed on the basis of the correlation b
etween the experimental values and descriptors which are derived
from the molecular structure of respective compounds
 QSPR are very useful techniques that are applied for the estimati
on of physicochemical and biological parameters for substances w
hich have not been examined by experiments.
 Usually, a QSPR model has the form of a mathematical equation
Property = f(x1, x2..., xn),

4. Introduction
 This equation relating particular topological, electronic, physicoc
hemical, etc. molecular descriptors x1, ..., xn to a property by m
eans of a certain function(f) of n variables.
 The function(f) may be an unknown, complex or non-linear.
 This approach assumes that there is strong correlations exist be
tween Structure and property of compounds considered . It is w
ell known that within the same group of organic compounds, th
at there are strong correlation between structure and observed
properties.
 For example, there is a relationship between the number of carb
ons in alkanes and their boiling points. Theoretical studies show
s that there is a clear trend in the increase of boiling point with
an increase in the number of carbons and this serves as a mean
s for predicting the boiling points of higher alkanes.
4

5. INTRODUCTION
 To obtain physical and chemical properties of these molecules. Chemical
structure of any drug determine its pharmacokinetics and
pharmacodynamic.
 Detail understanding of relationship between the drug chemical structure
and individual pharmacokinetic parameter is required for efficient
development of new drug.
 Different approaches have been developed for this purpose, ranging from
statistic based QSPR Aanalysis to physiologically based pharmacokinetic
models[PBPK].
 Nowadays, a large amount of experimental and predicted data about the 3D
structure of organic molecules and biomolecules is available.
 For drug discovery, it is very important It is very time consumming to
measure the properties. Therefore approaches for their prediction are a topic
of an intensive research.
 QSPR models are very strong tools for predicting these properties

6. 6
 One of key methodologies for processing these data is Quantitative
Structure-Property Relationship (QSPR) modeling.
 This methodology expresses molecules via various numerical values
(called descriptors), which encode the structural characteristics of
molecules.
 Afterwards the descriptors are employed to calculate the
physicochemical properties of the molecules.
 QSPR provides an effective way to estimate physicochemical properties
(e.g. dissociation constants, partition coefficients, solubility,
lipophilicity, biological activity, . . . ).
 The predecessors of QSPR models are the Quantitative Structure
Activity Relationship (QSAR) models, which are focused on estimating
of one particular property of a molecule – its biological activity.
WHY QSPR IT IS USEFUL :-

7. WHY QSPR IT IS USEFUL :-
 The chemical structure of any drug determines its pharmacokinetics and pharm
acodynamics. Detailed understanding of relationships between the drug chemi
cal structure and individual disposition pathways (i.e., distribution and elimina
tion) is required for efficient use of existing drugs and effective development o
f new drugs.
 Different approaches have been developed for this purpose, ranging from statis
tics-based quantitative structure-property (or structure-pharmacokinetic) relati
onships (QSPR) analysis to physiologically based pharmacokinetic (PBPK) m
odels.
 Models that can be used to predict different aspects of disposition are presente
d, including:
7

8. WHY QSPR IT IS USEFUL :-
(a) value of the individual pharmacokinetic parameter (e.g., clearance or volume
of distribution),
(b) efficiency of the specific disposition pathway (e.g., biliary drug excretion or
cytochrome P450 3A4 metabolism. presented pharmacological agents include
"classical" low-molecular-weight compounds, biopharmaceuticals
(c) accumulation in a specific organ or tissue (e.g., permeability of the placenta or
accumulation in the brain), and
(d) the whole-body disposition in the individual patients. Examples of presented
pharmacological agents include "classical" low-molecular-weight compounds,
biopharmaceuticals, and drugs encapsulated in specialized drug-delivery systems
8

9. QSPR [QUANTITATIVE STRUCTURE PROPERTY
RELATIONSHIP]
9
 The rapid development of modern computational technology has created an e
ntirely new environment for the efficient use of the theoretical construction in
many areas of applied research.
 The theoretical approach has proven to be especially beneficial in chemistry a
nd allied sciences, where the experimental-analysis and artificial developmen
t of new compounds and materials can be time-consuming, laborious, expensi
ve or even hazardous. Hence there is a need of alternate approach to experim
ental process.
 A feasible estimation method should satisfy the following conditions:
(i) It can be applicable to diverse set of compounds
(ii) It requires minimum number of inputs.
(iii) It should provide reasonable accuracy to the predicted output.
(iv) It needs minimum computation time.

10. QSPR [QUANTITATIVE STRUCTURE PROPERTY
RELATIONSHIP]
10
 Numerous estimation methods are reported in literature to predict a given physi
cal or chemical property. On the other hand, as the heterogeneity in the chemica
l structure of the compounds increases, property prediction becomes less reliabl
e and consumes more time.
 The different approaches generally used for prediction of physical properties ca
n be classified into the following categories
(1) Correlation based on experimental data.
(2) Group contribution method based on fragments.
(3) Correspondence theorem based on critical properties.
 QSPRs are computer-assisted mathematical models, which relate the physicoch
emical property of compounds to their chemical structure QSPR studies are per
formed on the basis of the correlation between the experimental values and des
criptors which are derived from the molecular structure of respective compound
s.

11. Models
 QSAR :- Quantitative Structure-Activity Relationship
A QSAR is mathematical relationship between a
Biological activity of a molecular system and its
geometrical and chemical characteristics
 PBPK :- Physiologically based Pharmacokinetics
Is a mathematical modeling technique for predicting the
Absorption,Distribution,Metabolism,Excretion (ADME)
 QSPR :- Quantitative Structure-Property Relationship
Property is related to Pharmacokinetics Parameter
same like PBPK Model
11

12. Advantages Of QSPR Model
 It can be applicable to diverse set of compounds.
 It requires minimum number of inputs.
 It should provide reasonable accuracy to the predicted output.
 It needs minimum computation time.
 Numerous estimation methods are reported in literature to pred
ict a given physical or chemical property.
12

13. Flowchart of the methodology used in our
QSPR Work
13

14. Flowchart of the methodology used in our
QSPR Work
14

15. Flowchart
15

16. Methods in QSPR
 QSPR studies consists of four stages:-
1. Selection of dataset and generation of
Molecular descriptor.
2. Descriptive analysis.
3. Statistical analysis. (prediction and evaluation
of model)
4. suggestion of novel compounds.
16

17. 1.Selection of dataset and generation of
Molecular descriptor
 In the first stage, the data sets of the propriety were collecte
d from previous works with known values of the studied effec
t and the values of descriptors were calculated.
 This is major step in the development of QSPR. is the generat
ion of molecular descriptors, which can describe the complete
molecular structure or any structural fragment.
 Molecular descriptors are numerical values that characterize t
he properties of molecules. In this work, VCCLAB based onlin
e software E-DRAGON is used to generate the descriptors. Ar
ound 1600 descriptors are calculated for each molecule.
17

18. Type of theoretical Descriptors with Examples
18
SR.NO Descriptor Type Meaning & Example
1. Topological
Descriptor
Describes Chemical bonds
e.g. Wiener index, Balaban
index, Randic indices,
Connectivity indices, Kappa
shape indices, Kier and Hall
indices.
2. Constitutional
Descriptors
Derived from atomic
composition.
e.g. Molecular weight,
Number of Individual type of
atoms, Number of atoms,
Number of Bonds, Number of
types of bonds.

19. Continue
19
3. Geometrical
Descriptors
Derived from 3-D structures
e.g. Surface area, Molecular
volume, Molecular steric
field).
4. Electronic
Descriptors
Charge distribution related
descriptors
e.g. Polarisability, Net atomic
charges, Dipole moment,
Hydrogen Bonding, Normal
modes.

20. Examples of drug Molecular descriptors that can
be used to
predict Drug Disposition
 Molecular descriptors
1. Size (MW) molecular weight
2. Shape
3. Molecular or polar surface area
4. logP (partition coefficient)
5. pKa (acid dissociation constant)
6. Specific functional groups
20

21. 2.Descriptive analysis
 All the descriptors generated for each molecule are not significa
nt in developing QSPR models. The use of all available descripto
rs in the model development process causes poor predictions be
cause of overfitting. Further, the use of redundant or irrelevant
descriptors diminishes the performance of a QSPR model.
 Descriptor selection is the process of identifying most relevant in
formation rich descriptors from large set of available descriptors.
 Several different methods are available in the literature for descr
iptor selection. The most widely used techniques are the Forwar
d selection.
21

22.  Technique:-
a) The Principal Component Analysis (PCA)
b) Hierarchical Cluster Analysis (HCA)
c) K-means Clustering methods
 Were used to form dissimilar clusters of compounds, to which the q
uery compounds would be compared for determination of degree of
similarity and the non-multicolinearity among variables (descriptors)
.
 After that, the dataset must divided into training and test sets.
22

23. a) Principal Component Analysis (PCA)
 PCA is a useful statistical technique for summarizing all the infor
mation encoded in the structures of the compounds.
 It is very helpful for understanding the distribution of the compo
unds.
 This is an essentially descriptive statistical method, which aims t
o present, in graphic form, the maximum of information contain
ed in the dataset compounds.
23

24. b) Hierarchical Cluster Analysis (HCA)
 The aim of the HCA was the recognition of groups of objects ba
sed on their similarity; it involves grouping a collection of object
s into clusters (subsets)
 such that objects within each cluster is more closely related to
one another than objects in different clusters.
 It is a multivariate chemo metric technique, which produced res
ult by class or cluster
24

25. c) K-means Clustering
 The k-means clustering is a non-hierarchical method of clusterin
g that can be used when the number of clusters present in the o
bjects or cases is known.
 In general, the k-means method will be produced exactly k diffe
rent clusters.
 The division of the dataset into training and test sets has be per
formed using the HCA or the K-means clustering technique.
 In this one, from each obtained cluster one compound for the t
raining set was selected randomly for used as test set compoun
d.
25

26. 3. Statistical analysis. (prediction and evaluation
of model)
 Before a QSPR model can be used to predict the physicochemica
l property of new molecules, an evaluation of the model is essen
tial.
 The models developed in this research are evaluated by Root M
ean Square Error(RMSE) and the squared correlation coefficient(
R2).
 The value of R2 is always between 0 and 1. The higher the valu
e of R2 represents more prediction accuracy of the model.
 The lower the value of RMSE of the predicted data represents h
igher prediction ability.
 In order to propose mathematical models and to evaluate quanti
tatively the physicochemical effects of the substituents on the pr
opriety.
26

27.  The developed models was compared and validated using
internal validation techniques, such as the key statistical t
erms (Correlation and determination coefficients r or r2) a
nd Leave one (or N) out Cross Validation CV-LOO/CV-LNO
methods.
 External validation using the test set(e.g.: group of molec
ules not in the original data training set on which the mod
el has been developed
27

28. Statistical Methods:-
 a) Multiple linear regression (MLR)
 b) Partial least squares (PLS)
 c) Multiple nonlinear regression (MNLR)
 d) Artificial Neural Networks (ANN)
28

29. a) Multiple linear regression (MLR)
 This method is one of the most popular methods of QSPR becau
se of its simplicity in operation, reproducibility and ability to allo
w easy interpretation of the features used.
 The important advantage of the linear regression analysis that a
re highly transparent, therefore, the algorithm is available and p
redictions can be made easily.
 Another advantage is that it can aid a priori descriptors selection
29

30. b) Partial least squares (PLS)
 PLS is a generalization of MLR, It can analyze data wi
th strongly collinear, correlated and noisy.
 If the number of descriptors gets too large (e.g., clos
e to the number of observations) in MLR, it is likely to
get a model that fits the sampled data perfectly in a
phenomenon called over fitting
30

31. c) Multiple nonlinear regression (MNLR)
 MNLR is a nonlinear method; in this one, we applied the descrip
tors proposed by the MLR corresponding to the dataset (training
set). In our previous works, we were used the preprogrammed f
unction:
 Y = a + (bX1 + cX2 + dX3 + eX4 + · · ·) + (f X12+ gX22+ hX3
2+ iX42+ · · ·)
 With: a, b, c, d ... represent the parameters and X1, X2,X3, X4...
represent the variables.
31

32. d) Artificial Neural Networks (ANN)
 To increase the probability of good characterization of studied c
ompounds, artificial neural networks (ANN) can be used to gene
rate predictive models of QSPR between the set of molecular de
scriptors obtained from the MLR, and observed activities.
 The ANN calculated activities model were developed using the
properties of several studied compounds.
 We were used the proposed a parameter ρ, leading to determin
e the number of hidden neurons, which plays a major role in det
ermining the best ANN architecture defined as follows:
32

33.  ρ = (Number of data points in the training set / Sum of the num
ber of connections in the ANN)
 In order to avoid over fitting or under fitting, it is recommended
that 1.8<ρ< 2.3. The output layer represents the calculated acti
vity/propriety values
33

34. Validation Techniques:-
 In order to assess the significance of the model and hence, its a
bility to predict proprieties of other (novel) compounds, the next
stage of the QSPR analysis consists of statistical validation:-
a) Internal Validation
b) External Validation
34

35. a) Internal Validation
 In this stage, our obtained models were validated internally by t
he cross validation techniques (such as the leave one out Cross
validation or the k-fold cross-validation…).
 In these techniques, the data will be partitioned firstly into k eq
ually sized segments or folds.
 One fold was eliminated from the data set, and the model was
then built using the remaining k-1 folds.
 The model thus formed was used to predict the activity/proprie
ty of the eliminated molecules .
 This process was repeated until all of the k folds.
 The cross-validation coefficient Q2(or R2cv) for the model was d
etermined based on the predictive ability of the model, the high
er value of Q2 (>0.5) indicate the better predictivity of the mod
el. 35

36. 36

37. b) External Validation
 The real predictive power of a QSPR model is to test their ability
to predict perfectly the activity/propriety of compounds from an
external test set (compounds not used for the model developme
nt).
 The purpose of a good QSPR model is not only to predict the ac
tivity of the training set compounds, but also to predict the activ
ities of external molecules (test set).
 This model will be able to predict the activity of test set molecul
es in agreement with the experimentally determined value.
37

38. b) External Validation
 The predictive capacity of the models that was judged, was bas
ed on the test validation coefficient R2test for the model determ
ined based on the predictive ability of the model for the test set.
 The higher value of R2test (>0.5) indicate the improved predicti
vity of the model
38

39. QSPR Approaches for Pharmacokinetic
Parameter (ADME)
 The effect of chemical structure on drug pharmacokinetics and on the r
esulting pharmacological effects has been based on empirical methods
for studying structure–Property relationships.
 Starting from the first half of the twentieth century, significant develop
ments have occurred in analytical chemistry, pharmacokinetics, pharma
codynamics, and other scientific fields, revealing the major mechanisms
that determine drug activity.
 The chemical structure of any drug determines its pharmacokinetics an
d pharmacodynamics. Detailed understanding of relationships between
the drug chemical structure and individual disposition pathways (i.e., di
stribution and elimination) is required for efficient use of existing drugs
and effective development of new drugs.
 Therefore, several approaches for investigating these processes have b
een developed, reflecting advances in the understanding of the pharma
cokinetic behavior of drugs.
39

40. Types of approach used to predict drug’s
Pharmacokinetics
40
Type of approach Examples of source data, in addition to the
drug molecular descriptors (physicochemical
properties)
In-silico (QSPR) Values of the individual pharmacokinetic
data/processes (e.g., drug clearance or brain
permeability
Cell-free in-vitro
systems
Drug retention on HPLC columns
Drug interaction with artificial membranes
Sub-cellular in vitro
systems
Drug metabolism by liver microsomes
Cellular in-vitro
systems
Drug permeation of cell monolayers
Drug accumulation in red blood cells

41. Types of approach used to predict drug’s
Pharmacokinetics
41
Ex-vivo
systems
Drug elimination by the perfused liver
In-vivo
experiments
Organ/tissue weights, permeability
coefficients , perfusion rates

42. Drug Absorption
 absorption is the sum of processes by which a drug proceeds fro
m the site of administration o the site from which the drug is tra
nsported to the site of action in the body.
 The most studied routes of absorption are the dermal and the g
astrointestinal route. In the present paper the discussion on abs
orption is based on the gastro intestinal route for which a large
body of in vivo data is available
 In an in vivo situation four rate-limiting steps can occur during a
bsorption: mucosal uptake, mucosal metabolism, gastric emptyi
ng, and blood flow.
 Mucosal uptake and metabolism create differences between lum
inal disappearance rates and blood appearance rates. The result
s of in situ techniques measuring disappearance rates should th
erefore be checked for eventual mucosal interaction
42

43. Drug Absorption
 Thus the real influence of lipophilicity on absorption may be blur
red by gastric emptying. Once the drug has passed the intestina
l membrane, it is carried away by the blood creating "sink condit
ions" which assure continuous absorption
 Highly lipophilic and small polar compounds penetrate so rapidly
through the membrane that the draining effect of blood flow be
comes the rate-limiting step for absorption
 The decline of the absorption rate for the higher members in a h
omologous series may therefore not be explained by simple part
ition models , but by a physiological limitation, e.g., mesenteric
blood flow
43

44. Drug Absorption
 The drug must pass the liver before reaching the systemic circul
ation. Indeed, virtually all blood perfusing the gastrointestinal tis
sues drains into the liver via the hepatic portal vein.
 The loss of drug occurring during the first passage of the gastroi
ntestinal membranes and liver is called the "first-pass effect." If
this phenomenon is not taken into consideration, false QSAR an
alysis will result, particularly if the metabolites produced are pha
rmacologically active.
 The first-pass loss can be assessed by comparing the pharmaco
kinetic data of oral administration with those following intraveno
us dose in which an initial passage of the liver is avoided
44

45. Drug Absorption
 Several methods allowing one to differentiate between pre abso
rptive, gut epithelial and hepatic first-pass biotransformation ha
ve been described.
 An important aspect in QSPR studies is the choice of pharmacok
inetic parameters for the description of a physiological process.
Some of the parameters are intercorrelated in complex ways, m
aking a non ambiguous interpretation difficult.
 Absorption of a series of homologs described by the peak plasm
a level (Cmax)or the area under the concentration-time curve (A
UC) values is often directly correlated with lipophilicity
45

46. Drug Absorption
 The problems due to the use of these parameters ha
ve been recognized by, the direct comparison of the b
lood level of chemical analogs fails to take into accou
nt that these parameters are the complex result of th
ree pharmacokinetic phenomena.
 Besides absorption, they contain the process of distri
bution and elimination [Equations. (1)-(5)].
46

47. 47

48. Drug Absorption
 where C is the plasma drug concentration (amount/volume),
 k is the elimination rate,
 ka is the absorption rate,
 F is fraction absorbed and
 D is dose thus
 FD is the amount of drug in plasma;
 V is volume of distribution;
 C1 is total plasma drug clearance,
 AUC is the surface under the concentration time profile; =
 tma is the time to obtain the maximal plasma drug concentratio
n Cmax-
48

49. Drug Absorption
 For the same amount absorbed (FD) and identical rate absorptio
n (ka) the increase in Cmax can be due either to a decrease of t
he volume of distribution (V), the elimination rate constant (k) o
r a combination of both.
 Consequently, only when V and k are constant, will the change i
n Cmax represent a change in ka or F.
 The separation of absorption and disposition (events following t
he absorption process) is possible with deconvolution procedure
s. The simplest of these procedures, permitting assessment of t
he absorption rate constant, is the graphical method known as t
he method of residuals. More sophisticated techniques have bee
n reviewed by Cutler
49

50. Drug Absorption
 Another frequently used, but inappropriate, descriptor of absorp
tion in QSPR is the percentage of drug absorbed (%abs). It can
be demonstrated that the relationship between %abs and log P
[Eq. 6] is not correct for extrapolation to high lipophilicities, for
which %abs becomes greater than 100%
log (%abs) = a log P + b.........(6)
 proposed for first-order absorption kinetics to transform %abs t
o ka by the expression
ka = -ln[1 -(%abs/100)]/t ……..(7)
 Drug absorption has been extensively studied using two- and th
ree-compartment models.
 Here we repeat only that these models revealed a bilinear relati
onship between drug absorption (log ka) and lipophilicity (log P)
. 50

51. QSPR study of maximum absorption
wavelength of various flavones
 A novel quantitative structure-property relationships (QSPR) mo
del has been developed for the maximum absorption wavelengt
h (λmax) of 69 flavones.
 Modeling of λmax of these compounds as a function of the bidi
mensional images as descriptors was established by chemo metr
ics methods.
 The resulted descriptors were subjected to principal component
analysis (PCA) and the most significant principal components (P
Cs) were extracted.
 Multivariate image analysis applied to QSPR modeling was done
by means of principal component-least squares support vector
machine (PC-LSSVM) method.
51

52. QSPR study of maximum absorption wavelength
of various flavones
 This model was applied for the prediction of the λmax of flavones,
which were not in the modeling procedure with low standard errors
and high correlation coefficient.
 The resulted model showed high prediction ability with root mean s
quare error of prediction of 0.3815 for PC-LSSVM.
 Chemical structure of 2-phenylchromen-4-one (2- phenyl-1-benzop
yran-4-one)
 λmax :- 250 nm
52

53. QSPR study of maximum absorption wavelength
of various flavones
53

54. QSPR study of maximum absorption wavelength
of various flavones
 This QSPR model exhibiting a high degree of accuracy was whe
n validated by predicting the λmax of experimental compounds i
n the external test.
 The results well illustrate the power of pixel descriptors in predic
tion of λmax of flavones.
 The work is the first application of MIA descriptors and PC-LSSV
M for QSPR study and shows that MIA descriptors are capable t
o recognize the physicochemical information and may be useful
to predict the maximum absorption wavelengths.
54

55. Different in Silico Models to Predict Human
Intestinal Absorption
55
Statistical
Method
Descriptors Database Performance of the
Best Model
MLRa Abraham descriptors Ni
tr= 31
Nii
test=138
Q1
Training = 85%
Q3
CV = 78%
MLR Sub structural molecular
descriptors
Ntr=417
Ntest= 50
QTraining = 79%
Q2
Test = 79%
LDAb TOPS-MODE descriptors Ntr= 82
Ntest=127
89 % of good
classification
93 % of good
classification
PLSd and
SVM
ADRIANA code, Cerius
index
Ntr = 380
Ntest =172
QTraining = 72-81%
QTest = 83-89%

56. Different in Silico Models to Predict Oral
Bioavailability
56
Statistica
l Method
Descriptors Database Performance of the Best
Model
MLRa 85 fragment Ni= 591 Q1
training= 0.71
Q3
CV (LOO)= 0.63
Q4
CV (LGO)= 0.58
(80/20)
MLR Physicochemical
properties, topological,
constitutional,geometrical
and quantum chemical
descriptors
Ntr= 159
Ntest= 10
Qtraining= 0.35
QCV (LOO)= 0.25
Q2
test= 0.72
GAd-
QSPR
Multiple molecular
descriptors
N= 577 Qtraining= 0.55
QCV (LGO)= 0.42
(90/10)

57. QSPR Approach for Drug Disposition
(Distribution & Elimination)
 The volume of distribution (Vd) is an important PK parameter th
at relates drug serum concentrations to the amount of drug in t
he body.
 The drug distribution in the body mainly depends upon plasma
protein binding and tissue binding.
 Vd has a significant impact on other PK properties, such as clear
ance and half life.
 Following figure determine Different levels of parameters affect
the pharmacokinetic behavior of the drug.
57

58. The pyramid of factors that determine drug
Disposition
58

59. QSPR Approach for Drug Disposition
(Distribution & Elimination)
 Three groups of the underlying factors are the physicochemical p
roperties of the drug, the physiological parameters of the body, a
nd features related to drug administration (level 0).
 The interplay of these variables determines the values of the 1st
level of pharmacokinetic parameters. For example, permeability c
oefficient (KP) of drug accumulation in a specific tissue is determ
ined by the drug size and/or lipophilicity and tissue composition
and/or perfusion. Values of the volume of distribution and cleara
nce (2nd level) reflect interplay of the underlying variables.
 For example, perfusion- or permeability-limited elimination of the
drug by the liver depends on liver perfusion, the extent of bindin
g to plasma proteins, and intrinsic clearance of the drug by the
metabolic systems of the liver.
59

60. QSPR Approach for Drug Disposition
(Distribution & Elimination)
 Drug’s volume of distribution and clearance govern its half-life, a
ffect its input function after pre-systemic administration, and det
ermine the time course of drug concentration (levels 3 and 4).
 The dotted line indicates pre-systemic first-pass metabolism of t
he drug, which can limit its systemic bioavailability (e.g., after or
al administration).
 The aim of this work was to establish QSPkR models to predict t
he volume of distribution values of drugs, using only theoreticall
y calculated molecular descriptors.
60

61. QSPR Approach for Drug Disposition
(Distribution & Elimination)
 For that a large set of descriptors was calculated and correlation
based feature selection (CFS) method was employed to select th
e best set of descriptors for modeling.
 In order to find nonlinear relationships between descriptors and
Vd values, we also used artificial neural network (ANN) and sup
port vector machine (SVM) method, and compared the linear m
odels derived by the traditional multiple linear regression (MLR)
method.
61

62. 62
COMPUTATIONAL MODELING IN DRUG
DISPOSITION
 Historically, drug discovery has focused almost exclusively on efficacy and selectivity
against the biological target.
 As a result, nearly half of drug candidates fail at phase II and phase III clinical trials
because of the undesirable drug pharmacokinetics properties, including absorption,
distribution, metabolism, excretion and toxicity (ADMET).
 The pressure to control the escalating cost of new drug development has changed the
paradigm since the mod-1990s.
 To reduce the attrition rate at more expensive later stages, in vitro evaluation of ADMET
properties in the early phase of drug discovery has widely adopted.
 Many high-throughput in vitro ADMET property screening assays have been developed
and applied successfully.
 Fueled by the ever-increasing computational power and significant advances of in silico
modeling algorithms, numerous computational programs that aim at modeling ADMET
properties have emerged.
 A comprehensive list of available commercial ADMET modeling software has been
provided till date.

63. 63
IN SILICO MODELING TARGET OF
DRUG DISPOSITION

64. Descriptors and log Vd values of training and
test set compounds
No. Drug
Name
Ia In GATS1e GATS5
e
HATS
8m
Psy8
0
HAr
Rc
logVd
1. Acyclovir 0 1 0.672 1.124 0.023 0 0 -0.022
2. Adefovir 1 0 1.312 1.159 0.16 0 2 -0.377
3. Aspirin 1 0 0.833 1.379 0.002 0 0 -0.824
64

65. Descriptors and log Vd values of training and
test set compounds
65
No. Drug
Name
Ia In GATS1e GATS5
e
HATS
8m
Psy8
0
HAr
Rc
logVd
4 Codeine 0 0 0.882 0.507 0.01 0 0 0.322
5 Doxepin 0 0 1.042 0.754 0.102 1 0 1.068
6 Isoniazid 0 1 0.697 0.75 0 0 1 -0.17
7 Metformin 0 0 1.8 1.8 0 0 0 0.389

66. Steps involved in Prediction of Human Volume
of Distribution Values for Drugs
1. Materials and methods
1.1 Data set
1.2 Molecular Descriptors
1.3 Descriptor selection and linear model generation
1.4 Artificial neural network
1.5 Support vector machines
1.6 Validation techniques and model performance
Evaluation
2. Results and discussion
2.1 Descriptor selection and linear model
2.2 ANN models
2.3 SVM models
2.4 Comparison of MLR, ANN and SVM models 66

67. Statistical results of different QSPkR
models
67
Model R R2 RMSE AFE
MLR(train) 0.884 0.782 0.254 1.709
ANN(train) 0.905 0.819 0.235 1.658
SVM(train) 0.891 0.794 0.245 1.681
MLR(train) 0.749 0.561 0.323 2.05
ANN(train) 0.788 0.621 0.317 2
SVM(train) 0.762 0.581 0.312 1.99

68. QSPR Approach for Drug Disposition
(Distribution & Elimination)
 The results obtained demonstrate that a QSPkR based predictio
n using theoretically calculated descriptors can lead to reasonabl
e predictions of human pharmacokinetics Vd values.
 The statistical analyses of the training data indicate the superior
ity of the ANN model over SVM and MLR on predictive ability an
d accuracy of prediction.
 The results from the study also suggest that the Sanderson elec
tronegativities, atomic mass, the number of heteroaromatic ring
in the molecule, antipsychotic drug like properties, and acidity of
the molecule play a key role in the Vd values.
 Thus, the proposed models provide some insights into structural
features for screening compounds for pharmacokinetic propertie
s in early drug development stage and help in reduction of anim
al experiments.
68

69. PREDICTION OF INDIVIDUAL DISPOSITION
PATHWAYS FOR LOW-MOLECULAR-WEIGHT DRUGS
69
 Biliary excretion is an important disposition pathway that is involved in
elimination and enterohepatic cycling of some drugs.
 Biliary excretion occurs predominantly via adenosine triphosphate (ATP)-
dependent efflux pumps, including organic anion transporters (OATPs), and its
efficiency is highly dependent on the chemical structure of the drug
(predominantly on the MW and lipophilicity) .
 The effects of drug structure and biliary excretion have recently been
analyzed by two research groups by use of an in-silico QSPR approach.
 Yang et al. developed equations based on molecular predictors (2D and 3D) to
predict biliary clearance and the percentage of the dose excreted in the bile of
rats and humans.
 It was found that the efficiency of biliary elimination depends on the charge of
the molecule
1 Biliary Excretion

70. 1 Biliary Excretion
70
 MW threshold values for biliary excretion of organic anions of 400 and 475 g/mol
were determined for rats and humans, respectively; cations or neutral compounds
were not characterized by statistically significant MW threshold values.
 The values predicted by the QSPR model for biliary clearance in humans fell within
the threefold error range of observed values, but the fraction of the dose excreted in
the bile was predicted much less accurately.
 Chen et al. investigated the correlation of cumulative biliary excretion (measured in
bile duct cannulated rats) and with 2D molecular descriptors of drug structure by use
of a QSPR model.
 On the basis of analysis of 56 compounds with MWs in the 320–708 g/mol range, a
quantitative equation that included seven molecular descriptors was developed and
validated.
 Molecular hydrophobicity is the most important molecular property affecting
cumulative biliary excretion (higher lipophilicity was associated with lower biliary
excretion) with additional effects of the polarity and size of a molecule.

71. 2. P-Glycoprotein Inhibition
71
 P-glycoprotein (Pgp) is an energy-dependent efflux pump that has important
effects on the bioavailability and disposition of many drugs.
 Pgp-dependent transport of a specific substrate molecule limits its oral bioava
ilability, reduces the extent of its body disposition (including its permeability
to the brain, disposition via the placenta, etc.) and enhances its hepatic and re
nal excretion.Therefore, inhibition of Pgp can have a profound effect on drug
pharmacokinetics.
 Chen et al analyzed the correlation between physicochemical properties of 1,
273 molecules and Pgp inhibition (data from previous in-vitro measurements
of Pgp inhibition) by use of recursive partitioning (RP) techniques and Bayes
ian categorization modeling.

72. 2. P-Glycoprotein Inhibition
72
 On the basis of molecular solubility, log D (the apparent partition coefficie
nt at pH 7.4), MW, and other molecular properties.
 Prediction accuracy was 81.7 % for the 973 compounds in the training set a
nd 81.2 % for the 300 compounds in the test set. However, the applied appr
oach was suitable for classification purposes only, and not for quantitative
analysis of the extent of Pgp inhibition (e.g., concentration producing 50 %
inhibition, IC50, values).
 A similar limitation applies also to other previously developed approaches
used to predict Pgp inhibition

73. 3. Cytochrome P450 3A Metabolism
73
 CYP3A is the most abundant CYP in the human intestine and liver that contri
butes to the metabolism of drugs and limits their oral bioavailability (for exa
mple cyclosporine, nifedipine, verapamil, etc.). Thus, prediction of CYP3Am
ediated metabolism can aid in prediction of drug elimination and bioavailabili
ty.
 Heikkinen investigated the intestinal metabolism of 20 CYP3A substrates by
use of the GastroPlus PBPK model.
 The ‘‘in-silico’’ approach tended to underestimate intestinal metabolism with
20 and 65 % of the compounds falling into the 2- to 5- and 5- to 10-fold error
range, respectively.
 On the other hand, intestinal permeability of 95 % of the analyzed compound
s fell into the twofold error range for the ‘‘in-vitro’’ approach (based on perm
eability coefficients obtained in Madin Darby Prediction of Drug Disposition
421 canine kidney (MDCK) cell culture).

74. 4 Pharmacokinetic Interactions
74
 Pharmacokinetic models can be used to predict drug–drug interactions (DD
Is) and, thus, the required adjustments of drug dosing. Specifically, the effect
of individual transport/elimination pathways on the time course of drug conc
entrations in the presence of other drugs can be predicted.
 Currently, prediction of DDIs is usually based on the outcomes of in-vitro m
easurements. For example, induction of CYP3A4 in clinical settings and its c
ontribution to human clinical DDIs has been predicted on the basis of in-vitr
o measurements of CYP3A4 induction in hepatocyte cell culture, plasma and
hepatocyte drug binding, and other parameters .
 Similarly, in-vitro models based on suspended hepatocytes, liver microsomes
, and sandwich-cultured hepatocytes have been used to determine the intrinsi
c clearance for 13 compounds and to predict human hepatic clearance and m
etabolism and transporter-based DDIs .

75. 4 Pharmacokinetic Interactions
75
 Complex interactions that take place in different organs and tissues can be ana
lyzed by use of these tools , taking into account metabolism and transporter ef
fects, and permeability.
 It can be stated that currently existing methods for prediction of individual dis
position pathways of low-MW drugs are characterized by low accuracy.
 Some of these methods are suitable for classification purposes only, and can o
nly partially suit the needs of the researchers in drug discovery and developme
nt.

76. 4 Pharmacokinetic Interactions
76
 Most probably, improved methods for prediction of individual disposition pat
hways will come from the field of PBPK modeling. These models have been i
ncreasingly used during drug development and regulatory review in predicting
the efficiency of the individual disposition pathways and their changes as a res
ult of DDIs .
 It should be noted that several currently available PBPK software packages, f
or example GastroPlusand Simcyp Simulator, incorporate molecular predict
ors for data input, can be used for assessment of the individual pharmacokinet
ic processes, and are suitable for prediction of the concentration time-courses
of the studied drugs and their changes because of DDIs.

77. 77
1.Blood–Brain Barrier Drug Penetration
Reliable estimation of drug permeation via the blood–brain barrier (BBB) is imp
ortant for design of drugs acting on the CNS, and for safety assessment of drugs a
cting elsewhere in the body.
The BBB is a complex structure and its permeation depends on the drugs’ physi
cochemical properties, and on transport by means of influx and efflux pumps, incl
uding Pgp, breast cancer resistance protein (BCRP), OATP, amino acid transport
systems, and others.
Several QSPR models have been proposed for analysis of drug permeation via t
he BBB and brain accumulation based on molecular descriptors ; these differ in t
heir structures, the data analyzed (e.g., brain-toplasma ratios, cumulative brain ac
cumulation, etc.), and predictive capabilities.
A QSPR model of passive transport via the BBB based on data from 178 drugs
was recently proposed
Prediction of the Distribution of Low-Molecular
Weight Drugs to Individual Organs and Tissues

78. 2 Permeability Of The Placenta To Drugs
78
 The ex-vivo human placental perfusion method is the most popular and relia
ble method for assessing placental transfer and metabolism(usually measure
d as the placental e-consuming, and dependent on the availability of placenta
e from suitable donors.
 Several in-vitro models have been developed for analysis of the permeability
to drugs of the placenta, including primary trophoblastic cells, immortal cell
lines of placental origin, placental explants, and others, but they only partiall
y reflect the active transport (influx and efflux transporter-mediated), metabo
lism, and tissue-binding mechanisms that occur in vivo.
 Several QSPR models have been developed for analysis of the dependence o
f the placental transfer (measured by use of the ex-vivo human placental perf
usion method) on the drugs’ chemical structure, and critical analysis of sever
al such models has been performed

79. 2 Permeability Of The Placenta To Drugs
79
 A simple model for prediction of milk/plasma (M/P) drug concentration ratios
on the basis of pKa, plasma protein binding, and octanol/water partition coeffic
ients has been applied and had good prediction characteristics for a set of 10 ba
sic drugs. However, this model provided unreliable Prediction of Drug Disposi
tion 423 predictions of M/P ratios for a set of 69 drugs with more diverse chem
ical properties (e.g., acidic, basic, and neutral compounds, etc.).
 Subsequently, Zhao et al developed an approach for prediction of M/P ratio cla
ssification (M/P ratio lower or higher than 0.1) based on a set of 126 drugs.
 The vector machine analysis method that resulted in *90 % classification accu
racy and identification of the five major classifying molecular descriptors, the
most important being the logP of the drug (higher logP values were associated
with lower M/P ratios).
 Unfortunately, this model is suitable for classification purposes only and does
not provide quantit

80. METABOLISM
80
 Drugs and other xenobiotics that gain access to the body may undergo 1 or mo
re of 4 distinct fates, as follows:-
1. Elimination unchanged
2. Retention unchanged
3. Spontaneous chemical transformation
4. Enzymic metabolism
 Each of these fates are of importance but, in quantitative terms it is enzymic m
etabolism, often also referred to as biotransformation, that predominates.
 The main site of metabolism of foreign compounds is the liver, although extrah
epatic tissues, frequently the site of entry to or excretion from the body (e.g., lu
ngs, kidneys, gastrointestinal mucosa),also play a role in the metabolism of xe
nobiotics.

81. METABOLISM
81
 Compounds eliminated unchanged are generally either
(a) highly polar such as strong carboxylic or sulfonic acids (e.g.,
sodium cromoglycate) or quaternary amines (e.g.,
pancuronium), which if absorbed are rapidly cleared into the
urine or bile.
(b) volatile and hence readily lost via the lungs.

82. Quantitative Structure-Pharmacokinetic
Relationships for the
Prediction of Renal Clearance in Humans
 Renal clearance (CLR), a major route of elimination for many dru
gs and drug metabolites, represents the net result of glomerular
filtration, active secretion and reabsorption, and passive reabsor
ption.
 The aim of this study was to develop quantitative structure phar
macokinetic relationships (QSPKR) to predict CL of drugs or dru
g-like compounds in humans.
 Step-wise multiple linear regression was used to construct QSPK
R models for training sets and their predictive performance was
evaluated using internal validation
 All qualified models were validated externally using test sets. QS
PKR models were also constructed for compounds in accordance
with their:-
82

83. Quantitative Structure-Pharmacokinetic
Relationships for the
Prediction of Renal Clearance in Humans
1) Net elimination pathways. (net secretion, extensive net secretion,
net reabsorption, and extensive net reabsorption)
2) Net elimination clearances. (net secretion clearance, CLSEC; or net
reabsorption clearance, CLREAB)
3) Ion status.
4) Substrate/inhibitor specificity for renal transporters.
We were able to predict
1) CLREAB (Q2 = 0.77) of all compounds undergoing net
reabsorption.
2) CLREAB (Q2 = 0.81) of all compounds undergoing extensive net
reabsorption.
3) CLR for substrates and/or inhibitors of OAT1/3 (Q2 = 0.81),
OCT2 (Q2 = 0.85),MRP2/4 (Q2 = 0.78), P-gp (Q2 = 0.71), and
MATE1/2K (Q2 = 0.81). 83

84. Quantitative Structure-Pharmacokinetic
Relationships for the
Prediction of Renal Clearance in Humans
 QSPKR models can be used to predict CLR of compounds that
1) undergo net reabsorption/extensive net reabsorption
2) are substrates and/or inhibitors of human renal transporters.
 QSPKR models to predict biliary clearance and percent of administe
red dose excreted unchanged into bile in rats and humans
 Varma et al.(2009) analyzed 391 compounds to relate their physico
chemical properties to renal clearance in humans.
 The findings suggested that CLR correlates positively with polar sur
face area, number of rotatable bonds, and sum of H-bond donors a
nd acceptors and negatively with cLog P and Log D7.4.
 Moreover, neutral compounds predominantly undergo net reabsorp
tion, whereas weak acids and bases undergo net secretion
84

85. Definition of molecular descriptors selected into
QSPKR models
85
Descriptor Definition
SaasC Sum of (aasC–) electro-topological states
SssCH2_acnt Count of (– CH –) electro-topological states
SaasC_acnt Count of (aasC–) electro-topological states
Gmax Maximum E-state value of an atom in a molecule
Hmin Minimum hydrogen E-state value of an atom in a molecule
Hmax Maximum hydrogen E-state value of an atom in a molecule

86. Application of QSPKR models for the prediction
of renal clearance of New Molecular Entity
(NMEs) in humans.
86

87. Comparison of observed and predicted renal clearance
values of compounds with overlapping
substrate/inhibitor specificity for the renal transporters
87
Compound Observed OAT1/3 MRP2/4 P-gp OCT2 MATE1/2K
Acyclovir 3.57 5.14 - - - 4.21
Adefovir 3.33 2.81 3.60 - - -
Cimetidine 7.90,0.89 6.80 - - 0.50 6.46
Dofetilide 3.12,0.49 - - - 0.45 1.79
Famotidine 4.42,0.64 4.25 - - 0.67 6.07
Quinidine -0.22 - - -0.12 -0.17 -
Tenofovir 2.70 4.76 3.83 - - -

88. Current Mathematical Methods Used in QSPR
Studies.
 Recently, the mathematical methods applied to the regression of QAS
R/QSPR models are developing very fast, and new methods, such as
Gene Expression Programming (GEP), Project Pursuit Regression (PP
R) and Local Lazy Regression (LLR) have appeared on the QSPR stag
e.
 At the same time, the earlier methods, including Multiple Linear Regr
ession (MLR),Partial Least Squares (PLS), Neural Networks (NN), Sup
port Vector Machine (SVM) and so on, are being upgraded to improve
their performance in QSPR studies.
88

89. Multiple Linear Regression (MLR)
 Some new methodologies based on MLR have been de
veloped and reported in recent papers aimed at improv
ing this technique. These methods include:-
1. Best Multiple Linear Regression.(BMLR)
2. Heuristic Method.(HM)
3.Genetic Algorithm based Multiple Linear Regression.
(GA-MLR)
4. Stepwise MLR.
5. Factor Analysis MLR.
89

90. Partial Least Squares (PLS)
 In the field of QSPR, PLS is famous for its application t
o CoMFA and CoMSIA. Recently, PLS has evolved by co
mbination with other mathematical methods to give be
tter performance in QSPR analyses. These evolved PLS’
, such as
1.Genetic Partial Least Squares (G/PLS)
2.Factor Analysis Partial Least Squares (FA-PLS)
3.Orthogonal Signal Correction Partial Least Squares
(OSC-PLS)
90

91. Neural Networks (NN)
 As an alternative to the fitting of data to an equation
and reporting the coefficients derived therefrom, neur
al networks are designed to process input information
and generate hidden models of the relationships.
 One advantage of neural networks is that they are n
aturally capable of modeling nonlinear systems.
 Disadvantages include a tendency to overfit the data,
and a significant level of difficulty in ascertaining whi
ch descriptors are most significant in the resulting mo
del.
91

92. Neural Networks (NN)
 In the recent QSPR studies, RBFNN and GRNN are th
e most frequently used ones among NN.
1. Radial Basis Function Neural Network (RBFNN)
2. General Regression Neural Network (GRNN)
92

93. Support Vector Machine (SVM)
 SVM, developed by Vapnik as a novel type of machine
learning method, is gaining popularity due to its many
attractive features and promising empirical performanc
e.
 New types of SVM are coming in on the stage of QSPR
, such as:-
1.Least Square Support Vector Machine (LS-SVM)
2.Grid Search Support Vector Machine (GS-SVM),
3.Potential Support Vector Machine (P-SVM)
4.Genetic Algorithms Support Vector Machine (GASVM).
93

94. Gene Expression Programming (GEP)
 Gene expression programming was invented by Ferreir
a in 1999 and was developed from genetic algorithms
and genetic programming (GP).
 GEP uses the same kind of diagram representation of
GP,but the entities evolved by GEP (expression trees)
are the expression of a genome.
 GEP is more simple than cellular gene progression.
 It mainly includes two sides:-
1. The GEP chromosomes, expression trees (ETs), and
the mapping mechanism.
2. Description of the GEP algorithm 94

95. Project Pursuit Regression (PPR)
 PPR, which was developed by Friedman and Stuetzle is a powerfu
l tool for seeking the interesting projections from high-dimension
al data into lower dimensional space by means of linear projectio
ns.
 Friedman and Stuetzle’s concept of PPR avoided many difficulties
experienced with other existing nonparametric regression proced
ures.
 It does not split the predictor space into two regions, thereby allo
wing, when necessary, more complex models. In addition, interac
tions of predictor variables are directly considered because linear
combinations of the predictors are modeled with general smooth
functions.
 Another significant property of PPR is that the results of each inte
raction can be depicted graphically. 95

96. Quantitative structure-property relationship
(QSPR) of thiazolidin-4-one derivitives as RTIs
of HIV virus
 Aim:-
The aim of this study is to build a quantitative structure
property relationship (QSPR) of 66 thiazolidin-4-one derivatives in
order to predict their log P.
 Introduction:-
Performing computational drug design is an important step
for their synthesis and properties characterizations.In this work,
quantitative structure-property relationship(QSPR)of 66
thiazolidin-4-one derivatives was examined in order to predict
their logP which is the most commonly used measure of
lipophilicity in chemical molecules. These group of compounds act
as non-nucleoside reversed transcriptase inhibitors of HIV.
96

97. Quantitative structure-property relationship
(QSPR) of thiazolidin-4-one derivitives as RTIs
of HIV virus
 Methods:-
Two different quantum mechanics approaches including HF
and DFT were applied for energy minimization of structures and
different classes of molecular descriptors including quantum
chemical descriptors were generated for prediction of their LogP.
Numbers of descriptors which showed high correlation with
each other were removed by MATLAB software.The model
between structures and their LogP was built for both methods
with performing Multiple Linear Regression (MLR) in Spss
package
97

98. Quantitative structure-property relationship
(QSPR) of thiazolidin-4-one derivitives as RTIs
of HIV virus
 Results:-
Statistical results and application of developed model to the
test set demonstrates that the DFT model is reliable with good
predictive accuracy.(R2cal = 0.90, R2cv = 0.88) The lack of significant
difference between the original and modeled values of logP reveals
the validity of the built model which was built with 2D and 3D
descriptors. The coefficients of model are statistically significant.
 Conclusion:-
QSPR models can be used to predict molecular properties such as
LogP. In this research, MLR model was built in order to correlate
structure of 66 compounds with their LogP. Molecules that were
optimized by DFT method showed better correlation than HF method
that indicates the accuracy of the built model with 2D and 3D
descriptors. 98

99. Software used in our QSPR development study
99
Drawing chemical structures Marvin Sketch , ACD/Chem Sketch,
ChemBioDraw
Generating 3D structures Gauss View 3.0 and ChemBio3D.
Calculating chemical descriptors Gaussian 03, Marvin Sketch 6.2,
ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009, SPSS statistics 20 and
Matlab R2009b

100. the BioPPSy software and Workflow
 The software is available for download from https://sourceforge.
net/projects/bioppsy/.
100

101. Conclusions
 In the present work, we carried out a comparative st
udy of the results that we have achieved during the l
ast three years on the use of statistical methods for t
he quantitative relationship study of the structure of
various compounds with properties of these compoun
ds.
 overall QSAR = f(QSARi, QSPRj, QSBRk, QSTRl )
 where i,j,k and l = 1, . . ., n. QSBR and QSTR stand f
or quantitative structure-biotransformation and struct
ure-toxicity relationships,respectively.
101

102. Conclusions
 Our work was developed as and the development of
our skills and working methods. Other works are still
bidding and accomplishment that we have taken into
account the mistakes.
 Establishing a simple QSPR model is difficult to give a
good guidance for screening drug design or for experi
ment. For the successful application of the developed
models in prediction for new compounds, rigorous val
idations will be used.
102

103. References
 [1] S. Chtita, M. Larif, M. Ghamali, M. Bouachrine and T. Lakhlifi, Quantitative s
tructure–activity relationship studies of di-benzo[a,d]cycloalkenimine derivative
s for non-competitive antagonists of N-methyl-d-aspartate based on density fun
ctional theory with electronic and topological descriptors, J. of Taibah Univ. for
Sci., 9(2015)143–154.http://dx.doi.org/10.1016/j.jtusci.2014.10.006.
 [2] S. Chtita, M. Larif, M. Ghamali, M. Bouachrine and T. Lakhlifi, QSAR Studies
of Toxicity Towards Monocytes with (1,3-benzothiazol-2-yl) amino-9-(10H)-acrid
inone Derivatives Using Electronic Descriptors, Orbital: Electron. J. Chem. 7 (2)
(2015) 176-184.
 Hansch, C., Muir, R. M., Fujita, T., Maloney, P. P., Geiger, F., and Streich, M. The
correlationo f biologicala ctivity of plant growthr egulatorsa nd chloromycetind
erivativesw ith Hammett constantsa nd partitionc oefficients.J . Am. Chem.S oc.
85:2 817- 2824 (1963). 2. Hansch, C., and Fujita, T. p-or-r Analysis. A method f
or the correlation of biological activity and chemical structure. J. Am. Chem. So
c. 86: 1616-1626 (1964).
103

104. References
 Benet LZ, Broccatelli F, and Oprea TI (2011) BDDCS applied to over 900 drug
s. AAPS J 13:519–547.
 Bolton EE, Wang Y, Thiessen PA, and Bryant SH (2008) PubChem: integrated
platform of small molecules and biological activities. Ann Reports Computatio
nal Chem 4:217–241.Doddareddy MR, Cho YS, Koh HY, Kim DH, and Pae AN (
2006) In silico renal clearance model using classical Volsurf approach. J Chem
Inf Model 46:1312–1320.
 Balant-Gorgia, A.E., Balant, L.P., Andreoli, A. 1993.Pharmacokinetic optimizati
on of the treatment of psychosis.Clin Pharmacokinetics 25, 217-236.
 Bauer, L.A. 2008. Applied Clinical Pharmacokinetics.2nd Edition, McGraw-Hill
Medical, New York.
 Berellini, G., Springer, C., Waters, N.J., Lombardo,F. 2009. In silico prediction
of volume of distribution
 Zupan, J. 1994. Introduction to artificial neural network(ANN) methods: What
they are and how to use them? Acta Chim Slov 41, 327-352.
104

105. 105