Kein Folientitel - PowerPoint PPT Presentation

1 / 39
About This Presentation
Title:

Kein Folientitel

Description:

(Quantitative Structure Activity Relationships) ... potential, , steric fields, hydrophobicity, solvent accessible surface area etc. ... – PowerPoint PPT presentation

Number of Views:84
Avg rating:3.0/5.0
Slides: 40
Provided by: hansb
Category:

less

Transcript and Presenter's Notes

Title: Kein Folientitel


1
Introduction to QSAR (Quantitative Structure
Activity Relationships)
2
Introduction to QSAR
  • Example of a Qualitative Structure Activity
    Relationship (SAR)

Affinity to Serotonin Receptor 1. decreases by
N-alkylation (R1/R2) 2. decreases by
bis-methylation (R3/R4) 3. increases by
methoxylation in R5 and/or R7 4. increases
by adding lipophilic substituents in R6
3
Introduction to QSAR
  • In contrast to Qualitative SAR, Quantitative SAR
    (QSAR) seeks to find a mathematical relationship
    between biological activity and molecular
    properties
  • General form of a QSAR equation
  • biol. activity f(P) with P molecular
    property/ies
  • or more specifically
  • biol. activity const. (c1.P1) (c2.P2)
    (c3.P3) ...
  • Molecular properties (descriptors) P are
    calculated for each molecule in the data set
  • Coefficients c and constant term are calculated
    by statistical methods (e.g. multiple linear
    regression)

4
Molecular QSAR-Descriptors
  • 1D Whole-molecule properties (e.g. molecular
    weight, melting point, logP etc.)
  • 2D Substituent constants (e.g. ?, ?, molar
    refractivity), fragment fingerprints,
    topological indices
  • 3D Surface or field properties (e.g.
    electrostatic potential, , steric fields,
    hydrophobicity, solvent accessible surface
    area etc.),

5
Introduction to QSAR
  • Why QSAR?
  • QSAR models are derived from a series of
    (similar) molecules with known activity (training
    set)
  • If a statistically relevant QSAR model has been
    found, it can be applied to new molecules in this
    series (test set) in order to predict their
    activity before biological testing (or even
    before synthesis!)

6
Introduction to QSAR
Example Analgesic activity of Capsaicin analogs
(taken from Walpole et al., Sandoz)
7
Introduction to QSAR
8
Introduction to QSAR
  • The Gibbs-Helmholtz equation (?GRTlnK) tells us
    that there is a logarithmic relationship between
    equilibrium constants (e.g. EC50) and free energy
    of binding
  • Thus, we have to transform the EC50 values to a
    logarithmic scale

9
Introduction to QSAR
  • Now, we require some molecular properties
    (descriptors)...
  • The Sandoz group decided to use two substituent
    constants the hydrophobic constant ? and the
    molar refractivity (MR) (correlated with the size
    and polarizability of the substituents)

10
Introduction to QSAR
  • We can plot the descriptor values vs. Log EC50 ...
  • ...and we can calculate linear equations for both
    parameters
  • Log EC50 0.76 - (0.82)?
  • Log EC50 1.14 - (0.07)MR

Our first QSAR equations!!!
11
Introduction to QSAR
  • How larger are the errors that we make?

?-Equation Log EC50 0.76 - (0.82)?
12
Introduction to QSAR
  • How larger are the errors that we make?

MR-Equation Log EC50 1.14 - (0.07)MR
13
Introduction to QSAR
  • How larger are the errors that we make?

Actual (measured) Log EC50 vs Predicted Log EC50
Correlation coefficients (R2) 0.88 0.53
14
Introduction to QSAR
  • Can we do any better by using both parameters in
    the equation (multiple linear regression (MLR)
    instead of simple linear regression)?

Best MLR Equation Log EC50 0.76 - (0.82)?
(0.0003)MR (Corr. Coeff. R2 0.89)
15
Introduction to QSAR
  • How can we validate these QSAR models?
  • Prediction within the training set (e.g. by
    leave-one-out cross validation)
  • leave out each compound once
  • calculate QSAR model with remaining compounds
    only
  • predict activity of left-out compound
  • compare prediction with "true" affinity
  • calculate "cross validated" R2 (often reported as
    Q2)
  • Prediction of the test set

16
Introduction to QSAR
  • How can we validate these QSAR models?
  • Prediction of the training set (cross
    validation)
  • Log EC50 0.76 - (0.82)? R20.88 Q20.72
  • Log EC50 1.14 - (0.07)MR R20.53 Q20.28
  • Log EC50 0.76 - (0.82)? (0.0003)MR R20.89
    Q20.58

17
Introduction to QSAR
  • How can we validate these QSAR models?
  • Prediction of the test set (compound 6i in this
    example)
  • Log EC50 0.76 - (0.82)? Predicted EC50
    for 6i 1.56
  • Log EC50 1.14 - (0.07)MR Predicted
    EC50 for 6i 0.42
  • Log EC50 0.76 - (0.82)? (0.0003)MR
    Predicted EC50 for 6i 1.57
  • Now we have a problem....

18
Introduction to QSAR
  • Some problems associated with "classical" QSAR
  • Only applicable within a chemical series
  • A good training set must be available
  • Activity data should be evenly spread
  • Activity data should span 3-4 orders of magnitude
    (log units)
  • Choice of meaningful descriptors
  • Problem of extrapolation (e.g. descriptors of
    test compounds lie out of descriptor range of
    training set)
  • Non-linear relationships are hard to detect

19
Artificial Neural Nets (ANN) An alternative way
of deriving QSAR models
20
J. Zupan, J. Gasteiger
Neural Networksin Chemistryand Drug Design
Second Edition
WILEY-VCH, Weinheim, 1999
21
Artificial Neural Nets
number of chemistry-related publications
927
855
743
441
498
290
105
30
3
5
22
Artificial Neural Nets
The "100 Steps Paradoxon"
human brain computer reaction time of firing
of neuron clock ratebasic unit 10-3 sec 10-9
sec (500 MHz) recognition of the faceof a
friend 10-1 sec 10-1 sec no. of processing
steps 100 100,000,000
  • The human brain works highly parallel

23
Artificial Neural Nets
Visual Cortexof the Human Brain
24
Artificial Neural Nets
Biological and Artificial Neurons
25
Artificial Neural Nets
Input e.g. molecular descriptors Output e.g.
biological activity
26
Artificial Neural Nets - Supervised learning
predict new compounds
change weights
27
Artificial Neural Nets
Net architecture ("topology") in Capsaicin
example
28
Artificial Neural Nets
Capsaicin dataset Predicted vs Actual Log EC50
using a trained Neural Net(Corr. Coeff. R2
0.92)
29
Artificial Neural Nets
  • Prediction of the test set (compound 6i)
  • (Multiple) linear regression predictions
  • Log EC50 0.76 - (0.82)? Pred. Log EC50 for 6i
    1.56
  • Log EC50 1.14 - (0.07)MR Pred. Log EC50 for
    6i 0.42
  • Log EC50 0.76 - (0.82)? (0.0003)MR Pred. Log
    EC50 for 6i 1.57
  • Neural Net prediction Pred. Log EC50 for 6i
    1.05
  • And the experimental affinity is......
  • Log EC50 gt3!!!(totally inactive)

30
Artificial Neural Nets - Unsupervised learning
?projection with preservation of the topology of
the input space
31
Artificial Neural Nets - Unsupervised learning
  • Molecules exhibit their activity via complex
    surface properties
  • In a Kohonen Neural Net, compounds with similar
    surfaces are placed in similar areas in 2D
    space

32
Artificial Neural Nets - Unsupervised learning
112 Dopamine and 60 Benzodiazepine Agonists in
the bulk of 8,323 structures of unknown activity
Kohonen map (40x30)
33
Artificial Neural Nets - Unsupervised learning
benzodiazepine 50
benzodiazepine 39
benzodiazepine 43
benzodiazepine 22
benzodiazepine 49
34
Artificial Neural Nets
Kohonen map unsupervised learning
multilayer network supervised learning
35
3D QSAR Linking QSAR with Molecular Modeling
36
3D QSAR
  • Two standard methods
  • CoMFA (Comparative Molecular Field Analysis,
    Cramer et al.)
  • QuaSAR (Vedani et al.)

37
3D QSAR - The CoMFA Approach
38
3D QSAR - The CoMFA Approach
For the capsaicin example, CoMFA predicted Log
EC50-0.21!
39
3D QSAR - The CoMFA Approach
Write a Comment
User Comments (0)
About PowerShow.com