Transcript Virtual Screening

Drug Design Process
Discovery Phase
Tripos Software

SYBYL & its modules
• SYBYL, Concord , MOLCAD, SiteId,
Advanced Computation, GASP, DISCOtech,
HQSAR, QSAR, AMPAC, Advanced CoMFA,
Distill, clogP/cMR, Biopolymer, Composer,
GeneFold, ProTable, MatchMaker, Leapfrog,
HiVol, CombiLibMaker/Legion, CLM 3D option,
DYANA, Dynamics, Selector, Triad, Mardigras,
FlexS, CScore, Confort, Stereoplex,
DiverseSolutions, Receptor, MM2/3, VolSurf,
FlexX, CombiFlexX, Unity Base, Unity 3D
SYBYL
Graphics
window
Text
port
Molecular Modelling &
Visualisation
SYBYL/Base
MOLCAD
Advanced
Computation
AMPAC
MM2 & MM3
Confort
MOLCAD
Advanced Computation
– creates molecular conformations
• using a variety of methods
• allows user defined constraints
• or constraints defined from other
molecules
• with all SYBYL force fields for
energy calculations
MM3 & AMPAC
– generating high quality molecular
structures
• AMPAC - very high quality (semiemprirical calculations)
• MM3 - high quality (advanced
molecular mechanics calculations)
– generation high quality structral data
• MM3 heats of formation,
vibrational spectra
Biomolecular Software
BioPolymer
Composer
GeneFold
MatchMaker
ProTable
SiteID
LeapFrog
Biopolymer

Structure building
– proteins, DNA, RNA, Carbohydrates

Structure editing
– conformations
• alpha-helices, beta-sheets
– sequence
• mutation, deletion, insertion, disulphide bonds, cyclisation




Sequence alignment
Structural alignment
Loop searching
Secondary structure
– prediction and assignment
Composer
– builds protein structure from sequence
• using homology modelling to model
proteins of unknown structure based
on known protein structures
• when >30% sequence identity exists
with known structure
ProTable

analyses protein structure
• homology models or experimental
(X-ray, NMR)
• uses molecular spreadsheet
interactively with graphics
ProTable view
Compute structure
quality data in MSS
And visualise it on
the actual structure
SiteId

– determines where on a protein a ligand
may bind
• using 2 methods
Grid method
– automated determination of cavities
– places protein in grid
– determines which grid points are deep
within protein
– clusters these points and presents them to
user
Dihydrofolate
Dihydrofolate
Reductase
Reductase
- with
--backbone
backbone
cancer
target
ribbon
ribbon
Dihydrofolate
Dihydrofolate
Reductase
Reductase
-anti
with
cavity
MTX
detected
ligand
Structure Activity Relationships
(QSAR) and ADME
QSAR with CoMFA, Advanced CoMFA, HQSAR,
clogP/CMR, Distill, VolSurf
QSAR with CoMFA
– computes specialised descriptors
• CoMFA, CoMSIA…..
– determine structural factors
responsible for molecular properties
– generate models to predict biological
activity
• or other molecular properties
– visualise QSAR models
QSAR - 3D QSAR - CoMFA


Comparative Molecular Field Analysis
Descriptors are field strengths around
molecules - electrostatic, steric, Hbond ..
PLS
pKi = A + B(D1) + C(D2) + ...
CoMFA - Interpretation
High Coefficient (important) lattice
points can be plotted around
More
negative charge
molecular
stuctures

Less steric bulk
*courtesy MDL
VolSurf

Just regular QSAR
– but uses specialised (ADME relevant descriptors)
descriptors with PLS or PCA

What are the descriptors?
– 72 descriptors - 5 classes
•
•
•
•
•

Size & Shape
Hydrophilic regions
Hydrophobic regions
INTEraction enerGY (intergy moments)
Mixed descriptors
Predefined models
– CACO2 permeability (A,D), skin permeability (A),
Blood-Brain barrier (D), LogP
A = Absorption, D=distribution
Pharmacophore Perception
DISCOtech, GASP, Receptor
DISCOtech

3D database queries
– and molecular alignments
H-bond Donor
Hydrophobe
H-bond Acceptor
x1,y1,z1
D2
x4,y4,z4
H-bond Donor
x3,y3,z3
D1
x2,y2,z2
D3
D4
x5,y5,z5
Hydrophobe
H-bond Acceptor
Distance query - pharmacophore
Spatial
pharmacophore
points
specified
Hitquery
molecule
from distances
LeadScreen
specified
by -interfeature
D1,
D2, ... by x,y,z
How does DISCOtech work?
Low E
Conformations
Molecular
Structures
Conformer
generation
Pharmacophore
Model
Clique
detection
Chemical Informatics

UNITY
CONCORD
SteroPlex
ChemEnlighten
AUSPYX
HiVol/HiStats

Molconn-Z





Unity
– Searches databases of chemical
structures to return molecules
matching the query
• 2D substructure searching
• similarity searching - those
fingerprints again
• 3D searching, verify, rigid, fully
flexible
Unity example similarity searching

Return all compounds in LeadScreen
(50,000 compounds) at least 75%
similar to
Search Takes 7 seconds and returns 8 compounds
Unity - flexible 3D searching

3D searching
– return molecules which can present a specific
arrangement of functionality (as defined in query)

2 ways to do this
– rigid 3D search of multiple conformations
• generate many conformations for all mols in database
• do rigid searching on all conformations of all
molecules
– fully flexible searching (Tweak algorithm)
• store 1 conformation per molecule in the database
• flex it on the fly to match the 3D query
• slower than rigid searching but more valid hits
Concord
– Generates 3D molecular structures
from 2D input
StereoPlex
• Creates 3D structures of all structurally feasible
stereoisomers
Virtual Screening
FlexX
CScore
CombiFlexX
FleS
FlexX
– docks molecules into protein binding
sites
– generates docked conformations
– generates docking scores
– assess complementarity between
receptor & ligand
CScore
– calculates scores for ligands in
protein binding site
• using a number of different
functions
G_score D_score PMF_score
FlexX
– computes the consensus between
different scoring functions
Virtual Screen
Database
Screen with FlexX
FlexX Virtual Screening Results
100
90
80
70
60
50
40
30
20
10
0
0
10
20
30
40
50
60
70
80
90 100
% compounds screened
FlexX docking & scoring.
2873 compounds with 68 known actives
- PGRD data
Visualization
64 CPUs
SGI™ Origin® 3800
Supercomputers
Data Array
Workstations
Virtual Screen Methods
Screen with DOCK
Database
Top 1%~10%
Screen with FlexX
>2 Million Comps
Lead Compounds
Molecular Diversity &
Combinatorial Chemistry
 Legion
 CombiLibMaker
 DiverseSolutions
 Selector
Legion/CombiLibMaker
– builds virtual combinatorial libraries
– have 2 modes of operation
• core + sidechains
• combine reagents
–
provide output for DiverseSolutions - lib
design
Legion

Combine reagents mode
R3
O
O
R1
R2
+
N
N
R4
R3
15 diketones
R2
R1
N
N
R4
31 hydrazines
465 products
Selector
– filters compounds
– calculates descriptors - valid ones
– compares databases for similar
molecules
– diversity selection - using a variety of
distance methods
Selector - Filtering
Filter the database
Excluded/included
compounds
Filtering criteria
Selector - Diversity Selection

Distance based
– must use distance based with high-dimensional
metrics
– example is 2D but reality is 1000D

e.g. Dissimilarity selection
Pick
most
different
compound
from
1st
Pick
most
different
compound
from
selected
Select
compound
at random
Compound
DiverseSolutions - Diversity
Selection

Cell based
– BCUT1
divide space into cells - pick a compound
BCUT2 form each
Compound