CADD

ADMET PROPERTIES
PRESENTED BY,
ADARSHA GOVINDA K
2ND SEM. Ph. Chemistry
NGSMIPS
CONTENTS
• Introduction
• Importance
• In silico ADMET models
Introduction
• ADMET refers to absorption,distribution, metabolism, excretion and
toxicity.
• Analysing ADMET properties before producing or synthesizing a new
drug is very important.
• It will minimize the risk of drug being incompetent.
• These ADMET properties have important role in designing of drug.
• Around 60% of failure in newer drugs are related to ADMET.
• Earlier days, ADMET tools were applied at the end of drug design.
• But, now these are considered from early stages itself, in order to
remove any risk of drug being rendered useless due to its ADMET.
Importance of determining ADMET
• Drug design is a multi step, time consuming and costly process
which requires large manpower.
• Undesired pharmacokinetic properties are analysed during
developmental stage.
• Early decision.
• Quick.
• Attrition in drug development.
• There are certain properties that make a drug different from other
compounds.
• Several pharmacokinetics properties are involved in this mechanism.
• Bioavailability depends on absorption and liver first-pass metabolism.
• The volume of its distribution, together with its clearance rate,
determines the half-life of a drug and therefore its dosage.
• To identify compounds that have good pharmacokinetic and toxicological profiles,
ADMET studies should be started as early as possible in the discovery process.
• New formats of in vitro ADMET assays serving early discovery have been
developed from traditional assays through protocol simplification and making use
of technological advancements.
• These models are promising as early screening tools for drug candidates and for
designing more successful combinatorial or in-silico libraries, where there is a very
small amount of the compound and a computational approach is then the only
option for collecting this information.
Insilico ADMET Model

1 • Data-based approach

• Structure based
2
approach
Data-based approaches

Data-based approaches include quantitative structure-

activity relationship (QSAR) and quantitative structure-
property relationship (QSPR).
These approaches use statistical tools to explore the
linear or nonlinear relationship between certain
structural descriptors and observed parameters of a
particular ADMET property.
To relate the target property to the descriptors, linear methods
(e.g., multiple linear regression and partial least squares) and non-
linear methods (e.g., feedforward artificial neural network) have
been applied for analysis.
Data-based methods are relatively simple and are applicable in
almost all ADME-Tox properties. Such models require minimal
computer power and have high throughputs of up to thousands or
millions of molecules per hour.
Structure-based approaches

Structure-based methods use three-dimensional structural

information and quantum mechanics to assess the interaction
potential between the small molecules studied with
macromolecules.
They require 3D structures of ligands and macromolecules , and
thus, require more computer power compared to the empirical
models resulting in a relatively lower throughput, varying from tens
to a few hundred compounds per hour.
Prediction of absorption properties of molecule

►Many factors influence the gastrointestinal tract

absorption of drugs, and would fundamentally be
classified into three categories:

Physicochemical factors Physiological factors Formulation factors

• pKa • Gastrointestinal PH • Drug particle size and

• Solubility • Gastric passage crystal form and
• Stability • Small and large intestine • Dosage form such as a
• Diffusivity transit time solution, tablet, capsule,
• Lipophilicity And • Active transport and suspension, emulsion,
• Salt Form efflux and gel, and modified release.
• Gut wall metabolism
 Almost all of the computational approaches currently used to
predict absorption are based on the assumption that absorption
is passive, and can be predicted from molecular descriptors of
the compound.
 The most widely used model for the prediction of passive intestinal
absorption is the Rule-of-five model introduced by Lipinski.
Log P > 5

Molecular weight > 500

The number of hydrogen bond (H-bond) donors

(counted as hydrogens attached to N or O atoms) > 5

The number of H-bond acceptors (approximated as

the number of N and O atoms) > 10
 In the modeling of the intestinal absorption of molecules, the
PSA is one of the most widely used descriptors.
 Several reports are available showing an experimental correlation
of PSA with the apparent permeability through a Caco-2
monolayer or Caco-2 cells and rat ileum .
 The PSAd values take into account the shape and flexibility of a
drug, while it has also been shown that the PSA calculated for a
representative single conformer performs as well as the PSAd
used to predict permeability .
 Wessel and co-workers have reported a neural network model to
predict percent human intestinal absorption (%HIA) .
 The descriptors they used were topological, electronic,
geometric, charged-partial surface area (CPSA) and other related
ones.
 These methods, however, are not applicable to a large number of
compounds, because estimations of PSA and CPSA require
conformational analysis and molecular orbital calculations, which are
computationally impracticable for a large set of compounds.
 Others also use PSA to predict oral absorption developing an artificial

neural network model to predict %HIA of compounds from their

molecular structure parameters.

 These parameters are the PSA, the fraction of polar molecular surface
area (FPSA, polar molecular surface area/ molecular surface area), the
sum of the net atomic charges of oxygen atoms (Q(O)), the sum of the
net atomic charges of nitrogen atoms with net negative atomic charges
(Q(N)), the sum of the net atomic charges of hydrogen atoms attached to
oxygen or nitrogen atoms (Q(H)), and the number of carboxyls (nCOOH).
 A methodology combining a genetic algorithm search with neural
network analysis applied to the modeling of Caco-2 cell apparent
permeability was developed by Di Fenza and co-workers .
 Several molecular descriptors of the compounds were calculated
and the optimal subsets were selected using a genetic algorithm.
 The selected descriptors were shown to possess physico-chemical
connotations which are in excellent accordance with the well-known
relevant molecular properties involved in the cellular membrane
permeation phenomenon: hydrophilicity, hydrogen bonding
propensity, hydrophobicity and molecular size.
 The predictive ability of the models, although rather good for a
preliminary study, is somewhat affected by the poor precision of
the experimental Caco-2 measurements.
 The generalization ability of one model was checked on an
external test set. The result obtained is of interesting practical
application and stresses that a successful model construction is
strictly dependent on the structural space representation of the
data set used for model development.

Hou and co-workers developed models to predict human
intestinal absorption by using the genetic function
approximation technique, and for a training set of 455 compounds.

The best prediction model contains four molecular descriptors:
1) topological polar surface area,
2) the predicted distribution coefficient at pH = 6.5,
3) the number of violations of the Lipinski’s rule-of-five, and
4) the square of the number of hydrogen bond donors.
The high quality of the classification model was validated by the
satisfactory predictions on the training set (correctly identifying 95.9%
of the compounds in the poor-absorption class and 96.1% of the
compounds in the good-absorption class) and on the test set
(correctly identifying 100% of the compounds in the poor-absorption
class and 96.8% of the compounds in the good-absorption class).
 Software packages are available commercially for predicting human
intestinal fraction absorption based on estimates of solubility and
intestinal permeability .
 GastroPlus simulates gastrointestinal absorption and
pharmacokinetics for drugs administered orally or intavenously in
humans and animals. Oral absorption simulation is based on an
advanced version of the compartmental absorption-transit model and
provides estimates of the fraction of the dose absorbed. Permeability
in each compartment is scaled according to the pH of that
compartment, the logP and the pKa values of the drug.
 QMPRPlus must be used for pure in silico predictions, which takes an
input file of multiple structures and generates estimates for lipophilicity
(logP), effective permeability, apparent permeability, diffusivity, and water
solubility;
 The estimates of QMPRPlus are derived from correlation models using a
variety of data from human and in vitro studies together with primary
molecular descriptors of chemical structures.
 iDEA simulates human physiology and accounts for regional
variations in intestinal permeability, solubility, surface area and
fluid movement.
 The system is based on the STELLA (Structural Thinking
Experimental Learning Laboratory with Animation) simulation
software, a physiologically based absorption model defining each
intestinal segment as a separate compartment.
 It is used for the description of fluid movement in the
gastrointestinal tract with a calculation of drug absorption in each
intestinal segment over time .
 The summation of the flux calculations in each segment gives the
total absorption rate.
 The absorption model is coupled with a physiological metabolism
model, which provides estimates for the rate and extent of first pass
metabolism in humans; the combined system allows the prediction
of bioavailability for a compound from in vitro data.
 Prediction of distribution properties of molecule


The distribution will depend on the structural and physicochemical
properties of the compound.

Most drugs exhibit a non-uniform distribution in the body with variations that
are largely determined by their ability to pass through membranes and their
lipid/water solubility.

Proteins exert a large influence on the distribution process because drugs can
bind to a variety of them such as

albumin (acidic drugs)

alpha1-acid glycoprotein (basic drugs)

lipoproteins (neutral and basic drugs)

Drug distribution can be estimated using tissue: plasma ratios or
the volume of a steady state distribution ( Vss ).

These methods are based on the assumption that there is a
passive diffusion between tissue compartments.

The identification of the tissue: plasma partition coefficients
( Kp ) needed for an initial prediction of the volume of steady state
distribution ( Vss ) of a drug in humans.

The initial estimation of the Vss of a new drug can normally be
based on only two Kp values, those of muscle and fat. The muscle
Kp can be used to represent all lean tissues, including the residual
and fat Kp can be used for the distribution of basic drugs.


Physiological information on tissue composition (lipid/water/protein
fraction), the blood composition (lipid/water/protein) and the blood
flowing into the tissues are often used to develop a partitioning
model.

Plasma protein binding is also taken into consideration, using
three simple measurements to predict the distribution volume of a
new compound:

human plasma protein binding,

experimental logD and

pKa.

QSPR techniques for the prediction of volume of distribution have
also been used.

Structural descriptors consisted of partitioning, quantum
mechanics, molecular mechanics, and connectivity parameters and
genetic algorithm or step-wise regression analyses were used in
the selection of the variable and model development.

The blood-brain barrier (BBB) is an important element in the
regulation of the internal environment of the brain. Drugs that
act in the CNS need to cross the BBB, in contrast, drugs with
a peripheral target may not cross the BBB to avoid CNS side
effects.


Early predictions logBB (logarithm value of brain to plasma
concentration ratio) involved classical QSAR approaches
using various physicochemical parameters such as

the octanol-water partition coefficient (logP) ,

molecular size descriptors and

solvation parameters.
• A combination of molecular descriptors was used to predict log BB
using lipophilicity, polarity, polarizability, and hydrogen bonding
parameters and partial least-squares statistics, and it was
demonstrated that topological and constitutional descriptors used in
partial least squares correlate well with experimental logBB data.
• There is a model using PSA as the only descriptor, calculating the
PSA from all low-energy conformations of a particular compound.

• The results showed an excellent correlation with the experimental

logBB data, for the protocol.

• However, it includes an extensive conformational analysis for each

molecule, which clearly prevents its application in a virtual high
throughput screening.
 Prediction of metabolic properties of molecule

• The metabolic fate of a compound depends on a number of

variables related to both
the chemical itself (chemical structure, physicochemical properties, etc.)
and
the biological system (enzyme and its environment)


A typical drug metabolism pathway is the oxidation of the
drug (phase I oxidation), followed by conjugation (phase II
conjugation).
Several methods have been developed mostly:

based on the knowledge of the structure and mechanism of the enzymes (protein structure; 3D-structure and accessibility of the

binding site; catalytic activity; mechanisms; specificity and regioselectivity),

OR

based on the physicochemical properties of the compound (molecular sites sensitive to oxidation or conjugation, structure of the

chemical, molecular surface properties, electronic structure, quantum mechanics properties, etc)

Due to the importance of cytochrome P450 in drug metabolism
many models have been developed to predict both substrates
and inhibitors.

3D models of cytochrome P450 using the x-ray structure of
bacterial cytochromes P450 as templates, were developed.

By docking some substrates into the active site of their
metabolizing enzymes it was found that substrates binding into the
active site of isoform CYP2D6 are mainly favoured by hydrogen
bonding and electrostatic interactions between the substrate
and two residues of the enzyme.

In contrast, van der Waals attraction interactions mainly
contribute to stabilize the complex involving both CYP1A2 and
CYP3A4.
 Computer systems to predict xenobiotic metabolism are
commercially available. These softwares mainly predict phase I
metabolism, although some of them also predict phase II
reactions.
 COMPACT (Computer-Optimized Molecular Parametric Analysis
of Chemical Toxicity) [ has modules that assess the ability of
xenobiotics to form enzyme substrate complexes and undergo
metabolic activation by the CYP1A and CYP2E subfamilies of
cytochromes P450.
 The COMPACT model uses sterical and electronic parameters.
Improved criteria for CYP1A and CYP2E1 substrate specificity
have recently been reported .

META is an expert system that can predict the sites of potential
enzymatic attack and the nature of the chemicals formed by such
metabolic transformations.

The program uses dictionaries of biotransformation operators
created by experts in the field of xenobiotic metabolism to represent
known metabolic pathways .

Activation of a biotransformation operator within the program is
based on the recognition of key functional groups within the
complete chemical structure.

The program evaluates stable metabolites and transform them
further build up a complete metabolic tree.

Processing continues in this way until the program is able to
generate water soluble metabolites that are to be excreted.
 MetabolExpert predicts the metabolism of exogenous compounds in
plants and mammals. The system is composed of biotransformation
databases, a knowledge-base and prediction tools .
 There are two types of predictions in MetabolExpert.
 In the first, the system tries to match basic biotransformations to the
compound structure.
 The second type of analysis is an extended prediction model in
which metabolites generated from basic transformations are
compared to transformations in ‘learned’ trees.
 Analysis of ADMET properties

The softwares used for the analysis of ADMET properties are as

follows
 QikProp – Schrӧdinger

 Swiss ADME

 ADMET Predictor
 QikProp-Schrӧdinger
QikProp predicts the widest variety of pharmaceutically relevant
properties - octanol/water and water/gas log Ps, log S, log BB,
overall CNS activity, Caco-2 and MDCK cell permeabilities, log
Khsa for human serum albumin binding, and log IC50 for HERG
K+-channel blockage - so that decisions about a molecule's
suitability can be made based on a thorough analysis.
Accurate ADME properties:
QikProp bases its predictions on the full 3D molecular structure;
unlike fragment-based approaches, QikProp can provide equally
accurate results in predicting properties for molecules with novel
scaffolds as for analogs of well-known drugs.
Lead generation:
 QikProp rapidly screens compound libraries for hits.
 QikProp identifies molecules with computed properties that fall
outside the normal range of known drugs, making it simple to
filter out candidates with unsuitable ADME properties.
Lead optimization:
QikProp can play an important role during lead optimization by
analyzing similarity within a class of compounds as well as by
identifying compounds to avoid because they exhibit extreme
values of predicted properties.