Protein Modeling with Discovery Studio

Kimmo Mattila
[email protected]

          
 Discovery Studio

 Discovery Studio (DS) is a commercial molecular modeling

program for biological macromolecules (proteins, nucleic acid).

 World is full of good molecular modeling programs like:

• Sybyl
• Maestro
• gOpenMol
• VMD
• PyMol
• Discovery Studio

Is Discovery Studio any better?

          
 Discovery Studio at CSC

 CSC has a national academic license and installation

files for DS

 The program can be installed in users own Windows

or Linux pc (requires fixed IP-address)

 You can use DS in hippu1.csc.fi using X-term,

Nomachine or DS-client connection

 The token license system does not limit the amount of

installations but the amount of simultaneous users.
(Close your DS if you are not using it)

          
 Installing Discovery Studio

 Academic university researchers can instal the software to a local

computer
 Installation instuctions can be found form:
http://www.csc.fi/english/research/sciences/bioscience/programs/ds/ds_install
 A group vise license contract is needed
 A fixed IP address is needed
 Installation files are downloaded from the scientist's user interface
(CSC user account is needed)
https://sui.csc.fi/group/sui/downloads
 DS client only: 300 MB (windows), 400 MB (Linux)
 DS complete: 3,2 GB (windows), 3,7 GB (linux)

          
 Structure of Discovery Studio
• Both client and server are normally installed in the
same machine
• separate server machines can be used too
• In most cases you do not need to worry about this
• If you use Hippu1 as your DS server you do not
need to install the whole package but just the client

DS
DS Client DS Server
• Pipe Line Pilot
Default
ports: • Apache
•Interface
9943
•Visualization 9944 • Protocols (BLAST,
CHARMM, Modeler,
•Commands and tools CDOCKER etc. )

          
 Discovery Studio interface
Command menus
Toolbars

Protocols

Tools

3D Window
Hierarchy view (graphics)

          
 Discovery Studio menu commands
 File menu: Contains commands for tasks such as opening molecular data
files, saving files to disk, printing, and accessing windows.
 Edit menu: Contains commands for tasks such as copying and pasting,
selecting, finding, and setting preferences.
 View menu: Contains commands for tasks such as changing the way
objects appear in the various views and for choosing which views should
be shown or hidden.
 Chemistry menu: Contains commands for tasks that modify the chemical
makeup of the molecules.
 Structure menu: Contains commands for tasks such as adding or
removing labels, adding or removing structure monitors, calculating the
solvent accessibility, cleaning up geometry, and superimposing multiple
molecules.
 Sequence menu: Contains submenus and commands to manage protein
sequences and protein sequence alignments.
 Window menu: Contains commands that allow you to control the display
of open windows in the current Discovery Studio session.
 Help menu: Contains commands to access the Discovery Studio Help
system and the Accelrys website.

          
 DS toolbars

 “Toolbars” are mostly shortcuts to menu

commands
 However, some functions are used only
through Toolbars

 Not all toolbars are normally visible. You can

add or hide toolbars from:
View | Toolbars

          
 DS tools
 “Tools” contain methods to analyze
and modify your molecular model
 Tools panel can be made visible
from:
View | Explorers | Tools
 The CSC license covers most but
not all the tools
 Most of the tools are run within the
client but some require connection to
the DS server

          
 DS Protocols

 Protocols are more advanced modeling

and analysis tasks that are computed by
the DS server

 Tools panel can be made visible from:

View | Explorers | Protocols

 Note that the license of CSC does not

cover all the protocols and tools

          
 Hierarchy panel

 The hierarchy panel is opened form:

View | Hierarchy
 You can use hierarchy to select, atoms,
amino acids or molecules
 Selections can be made for the 3D-view,
Data table and the sequence window too.

 The logic in Discovery studio is:

1. First select the target
2. Then select the command

          
 Data table

 The Data table is opened form: View | Data table

 Data table shows data and values associated to your molecular model
 The data can be viewed, modified and sorted in the data table
 If value is in gray background it can’t be modified
 Try right clicking the data table window (you can find a hidden
command menus, that open by right-click, all over Discovery Studio)

          
 Help
 Discovery studio contains a large help system. No
WWW or printed manual is available DS interface
 Open help form Help | Topics (if search tools are not
visible in the Help window, press CONTROL + s)

Other tricks to find the missing functionality or parameter:

 Use the Feature search toolbar:

 Try right-click to the window or object
 Check, if your parameter is locates in preferences
Edit | preferences
          
 Proteins and PDB

          
 Using PDB data in DS

 http://www.rcsb.org/
 experimentally determined protein structures are
stored into PDB (Protein Data Bank) database
 Sources: X-ray diffraction (about 80%), NMR (15 %),
others (5 %)
 over 65 000 structures (many of them related and
nearly similar however)
 this is much less than the amount of known protein
sequences (UniprotKB contains over 10 million
sequences)

          
 PDBe Database

 “Processed” version of pdb

 Several search approaches e.g. ligand
search, interface analysis,
 PISA to study assemblies interfaces and
monomers

 http://www.ebi.ac.uk/pdbe/

          
 Using PDB data in Discovery Studio

 Protein structures can be automatically retrieved

from the PDB database server using the four letter
PDB-code.
• File | Open URL | PDB ID

 You can do sequence based similarity searchers to

PDB with BLAST protocols.
 Protocol: RCSB Structure search enables metadata
and sequece motif based searches.
 Local PDB formatted files can be imported too (e.g.
files retrieved form PISA database).
          
 Using PDB files

 Note that there can be several thighs that may need editing in
your PDB file before you can start to use modeling tools

 Things that often require editing in PDB files:

hydrogens are missing
some side chains are missing
part of main chain is missing
ligand structure is not recognized
multiple conformations for some side chains
several structures in the PDB file

 Other things to check in a PDB-file:

main chain omega, phi and psi angles
side chain rotamers
          
 Checking and fixing a protein

 Protein Reports and Utilities tool can be used to get an overall

view to the selected PDB entry

 Validate protein structure tool can be used to check the PDB

structure and report the problematic sites

 Clean function in the Protein Reports and Utilities tool can be

used to fix some of the errors automatically
 Hydrogens can be added by Chemistry/H
 Protonation states can be fix with Build and Edit Protein Tool.
 The protonation states of neutral histidines are selected using
Edit | Preferences | Protein Utilities

          
 Checking and fixing a protein

 Protocol: Protocols | General Purpose | Prepare Protein

• Standardize atom names, insert missing atoms in residues and remove
alternate conformations.
• Remove water and ligand molecules depending on the settings.
• Insert missing loop regions based on either SEQRES data or user
specified loop definitions.
• Optimize short and medium size loop regions with the LOOPER
algorithm (optional).
• Minimize the remaining loop regions (optional).
• Calculate the pK and protonate the structure (optional).

          
 PDB file format

 PDB-file contains only information about atom

locations (X,Y,Z).
 Data about bonds, partial charges or force field types
is not included
 The fourth column contains normally R-factor, but
can be something else too
 Common format for molecular data
 X-ray structures lack hydrogens and may contain
several copies of the protein.
 NMR-structures contain several overlapping
structures

          
 PDB file format
HEADER RETINOL TRANSPORT 27-JUL-92 1BRP 1BRP 2
COMPND RETINOL BINDING PROTEIN (HOLO FORM) 1BRP 3
SOURCE HUMAN (HOMO SAPIENS) PLASMA
…
HELIX 1 1 VAL 6 SER 8 4 ONE SHORT TURN 1BRP 71
HELIX 2 2 PRO 146 GLU 158 1 1BRP 72
SHEET 1 S1 9 GLY 22 LYS 30 0 1BRP 73
…
SEQRES 1 A 182 GLU ARG ASP CYS ARG VAL SER SER PHE ARG VAL LYS GLU
SEQRES 2 A 182 ASN PHE ASP LYS ALA ARG PHE SER GLY THR TRP TYR ALA
SEQRES 3 A 182 MET ALA LYS LYS ASP PRO GLU GLY LEU PHE LEU GLN ASP
…
ATOM 1 N GLU 1 22.826 21.377 -30.151 1.00100.00 1 1BRP 99
ATOM 2 CA GLU 1 23.744 21.686 -29.074 1.00100.00 1 1BRP 100
ATOM 3 C GLU 1 23.395 23.023 -28.464 1.00100.00 1 1BRP 101
ATOM 4 O GLU 1 22.798 23.102 -27.389 1.00100.00 1 1BRP 102
ATOM 5 CB GLU 1 25.225 21.681 -29.508 1.00100.00 1 1BRP 103
ATOM 6 CG GLU 1 26.155 20.992 -28.489 1.00100.00 1 1BRP 104
ATOM 7 CD GLU 1 27.285 21.840 -27.971 1.00100.00 1 1BRP 105
ATOM 8 OE1 GLU 1 28.301 22.075 -28.603 1.00100.00 1 1BRP 106
ATOM 9 OE2 GLU 1 27.087 22.244 -26.741 1.00100.00 1 1BRP 107
ATOM 10 N ARG 2 23.771 24.073 -29.182 1.00100.00 1 1BRP 108
ATOM 11 CA ARG 2 23.485 25.397 -28.690 1.00 86.16 1 1BRP 109
ATOM 12 C ARG 2 22.026 25.784 -28.629 1.00100.00 1 1BRP 110
…
HETATM 1483 O HOH 229 2.848 65.969 -30.833 1.00 53.03 1BRP1581
HETATM 1484 O HOH 230 38.756 38.831 -49.928 1.00 75.69 1BRP1582

          
 Forcefield methods

          
 Force field methods

 Quantum mechanics  Molecular mechanics

• electronic structure • atomic level
calculations • simple models of the
• Ab initio methods interactions to calculate
• Semi-empirical the energy of molecule
methods as a function of nuclear
• time-consuming positions only
• high level accuracy • Not so accurate as QM
methods
• < 500 heavy atoms
• >10000 atoms

          
 Force Filed methods….

 Force field methods are usually used for biomolecules.

 to study complex system (i.e. the binding site in protein),

quantum mechanics and molecular mechanics methods
can be combined (hybrid QM/MM).

          
 Force Filed methods….

 Advantage  Limitations
• large systems in • no atom level
reasonable electrostatics.
calculation time.
• dependent on the
• in some cases FF
can provide results quality and
as accurate as the availability of
highest QM in a parameters.
fraction of the CPU- • The calculated
time. energy is relative.

          
 General form for Forcefield

 Energy consists of sum of terms each describing the energy

required for distorting a molecule

• Ebond energy function for stretching a bond between two atoms

• Eangle energy required for bending an angle
• Etorsion torsional energy for rotation around a bond
• Eelec electrostatic energy ( non-bonded interactions due to distribution of the
electrons)
• Evdw Van der Waals energy (repulsion or attraction between non-bonded atoms)

          
 Example. Functional form of CHARMm force field:

bond length angle dihedral angle improper torsion

electrostatic van der Vaals

          
 Components of forcefield

 The forcefield contains the necessary building

blocks for the calculations of energy and force:
• A list of atom types.
• A list of atomic charges (if not included in the
atom-type information).
• Functional forms for the components of the
energy expression.
• Parameters for the function terms.

          
 Forcefield….

 Torsion and non-bonded energy terms are

most important for biomolecules.
 A force field is transferable (set of
parameters developed on a small number of
cases can be applied to much wider range of
systems).
 Force fields are used in molecular
mechanics and molecular dynamics
calculations.

          
 Forcefield functions and parameters

Goal: a simple function for reproducing

structural properties.
1) Empirical fitted force field: a functional form
and parameters is designed to satisfy
experimental results. ( cvff)
2) Ab initio fitted Force field: a functional form
and parameters are specified using
theoretical models and calculations. (cff)

          
 Parameter assignment

 the forcefield has the same functional form

for all atoms but different parameters for
each atom types.
 the atom type and its parameter depends on
how that atom is bonded (example
CHARMm has 38 different carbon types).
 atom type ≠ atom name

          
 Parameter assignment...

 molecule can be neutral but the charge

distribution is not equal=> partial atomic
charges are determined.

 Partial charges are important!

• hydrogen bonds
• ionic bonds
• dipole moment

          
 Isoleucine:

Elements Atom names CHARMm Partial

types charges

          
 How to use forcefield ?

 several forcefields are available commercially.

 the validation of the force field depends on for which
purpose it is designed and what properties are
studied.
 the quality of force field parameters is essential.
 the complexity of functional form
 computational power (The computational time for
calculating the force field energy grows as the square
of the number of atoms).

          
  The ability to perform a calculation is no guarantee
that results can be trusted !
• unsuitable forcefield gives wrong results.
 A common problem: a lack of (good) parameters.
 different forcefields cannot usually be merged but
the results can be compared.
 forcefield methods are good for predicting properties
for classes of molecules where a lot of information
exists.

          
 Forcefields in DS

Discovery Studio can use CHARMM forcefields:

CHARMm, CHARMm polar H, CHARMm19, CHARMm22,
CHARMm27, XPLOLIG, MMFF, cff

Location of forcefield files in Disocvery Studio

DiscoveryStudio17/share/forcefield

The InsightII manual chapter “Forcefield based simulations” gives a

good introduction to force fields and their applications:

https://extras.csc.fi/msimanual/doc/insight2005/ffbs/FF_SimulTOC.html

          
 Applications of forcefield methods

 Applications of forecefields methods

• Molecular mechanics= Minimisation= Optimisation
• Molecular dynamics=Simulation

 Basic assumption for using forcefield methods:

A REAL MOLECULE IS IN A STATE WHICH
CORRESPONDS A MODEL NEAR
POTENTIALENERGY MINIMUM

          
 Search strategies

 Several different strategies in use.

 all methods are not using plain atomistic
models
 Molecular dynamics
 Monte Carlo
 Genetic algorithm
 Fragment based method
 Point complementarity methods

          
 Energy minimization

 “A REAL MOLECULE IS IN A STATE

WHICH CORRESPONDS A MODEL NEAR
POTENTIALENERGY MINIMUM” ,
 a search for the minimum of the potential
energy surface defined by energy function, is
done.
 Minimum energy arrangements of the atoms
corresponds to stable state of the molecule.

          
 Three major protocols for minimization

 Steepest Descent
 Conjugate Gradient
 Adopted Bases Newton-Rhapson
 Powell

          
 The steepest descents method

 the gradient of potential energy determines

the direction which leads to a largest
reduction in energy. A step will be taken to
that direction.
 Robust and simple method,
 useful when starting far from minimum
 Convergence is slow near minimum

          
 Conjugate gradient method

 the gradient of the previous step is included

 more efficient convergence to minimum →number
of iterations smaller than for steepest descents
 works quickly when the molecule's structure is far
from an energy minimum.
 the best choice for general use.
 more complex →time per iteration is longer than
for steepest descents.

          
 Newton-Rhapson method

 use not only the first derivatives (i.e.the gradients)

but also the second derivatives ( the curvature of the
function) to locate a minimum.
 initial guess of structure need to be close to
minimum.
 Convergence is fast near a minimum
 Computationally demanding for systems with many
atoms (suitable for less than 100 atoms).

          
 Comparison of minimization methods

 what determines which method is best?

1) size of system
2) current state of optimization
 robustness (i.e. ability to reach minimum regardless of initial
conditions)
steepest descents > conjugate gradient and Newton-Raphson
 number of iterations
steepest descents > conjugate gradient > Newton-Raphson

          
 Strategy in minimization

 use steepest descents for first 10-100 steps to

remove bad contacts.

 then use Newton-Raphson or conjugate gradients

to complete minimization to convergence

 DS Smart minimizer:
1. steepest descents (max 1000 iterations)
2. conjugate gradient (max 1000 iterations)

          
 Strategy in minimization

 as a minimum approaches, the rate of

convergence slows down and minimisation
method crawls toward minimum at an ever
decreasing speed.
• Convergence criteria (either for energy or
conformation changes) and amount of
minimization cycles are used to end minimisation.

 The step size in minimisation algorithm is defined

either by energy change in pervious step or by
line search method.

          
 Local or global energy minimum?

 optimization methods can only locate the “close by”

minimum, which is normally a local minimum, not a the
global minimum from a given set of coordinates.

 To check if the minimum is the local or the global, all

conformations need to be searched and the number of
the minima grows typically exponentially with the
number of variables.

 Systematic search is not possible for complete proteins

and thus global minimum can not be defined.

          
 Local or global energy minimum?

 Systematic search can be N Number of Time

used to check possible possible
conformations, but only for
conformations
small systems.
1 3 3s
 table1: possible
5 243 4 min
conformation for linear
alkanes CH3(CH2)n+1CH3)
15 14348907 166d
1
F. Jensen: Introduction to Computational
Chemistry

          
 Applications of energy minimization

 relieve any unfavourable interactions in the initial

configuration.

 calculations of the energy of molecular structure

 conformational search procedures

 normal mode analysis

          
 Molecular dynamics

• Movements of molecule are simulated in certain temperature

• Temperature = thermal movement
• Each atom has position, mass and velocity
• Force field affects to the model according to the Newtons’ law
F=ma
or
2
d ri
mi 2 = -? i [E(r1 ,r2 ,...,rN )] i = 1,...,N
dt

          
 Molecular dynamics

 Total energy of the system is

• Etot = Epot + Ekin

 Normal dynamics (Verlet/leap frog)

 Langevin dynamics

          
 Using molecular dynamics

 Proteins are not rigid but flexible objects

• Dynamical models are sometimes needed
• fluctuations
• Conformational search and changes
• ligand binding
• estimating thermodynamical parameters

 Accuracy, size and timescales of the simulations

are quite limited (tens of thousands of atoms,
hundreds of nanoseconds)

          
 Molecular dynamics parameters

 Handling of nonbonded interactions

• All interactions can’t be explicitly included
• Cutoff
• Ewald summation
• Cell multipole

 Time step
• is limited by the highest frequency in the model
• 0,5-5 fs => millions of simulation steps are needed
to reach nano second time scale.

          
 Molecular dynamics parameters

 Temperature
• normally 300 K
• Thermal equilibrium requires temperature control
or careful heating
• Simulated annealing utilizes higher temperatures

 Other parameters
• Force field scaling
• dielectricity

          
 Solvent environment
 Solvent (water) environment requires more
computing but makes model more realistic
 Periodic boundary conditions are used to create
continuous solvent environment
 Discovery Studio: Protocol Simulation/Solvation

          
 Analysis of molecular dynamics

 The course of the simulation is recorded

(trajectory).
 Later on several properties can be analyzed
• Geometry (angles distances)
• Changes and fluctuations
• Energies
• Interactions, hydrogen bonds
• Distributions and correlations
 Discovery Studio: Animation toolbar , Analyze
trajectory tool and Analysis Protocols modules.

          
 Docking
 many protein related biological processes are
regulated or enabled by specific binding of small
organic molecules (ligands) to the proteins
• signal transduction
• enzyme activity /inhibition

 many drugs are known or taught to work by

binding a target protein

 what molecules could bind to the active site or

where and how do the active molecules bind?

          
 Docking

 Two basic components:

1. scoring function
2. search strategy
initial position
optimizing search

 Systematic search is normally not possible

          
 Docking

 computer based docking can be used to predict binding

geometries for large libraries of candidate molecules, if
the protein structure is available
 speed is an issue (maximum duration few
minutes/ligand)
 What do we want to know?
• does this molecule bind or not?
• which of these molecules are most potential ligands?
• what is the binding geometry or site of this molecule?
• what is the binding affinity of this molecule?

 docking does not try to simulate the binding process!

          
 Scoring function

 scoring function should distinguish the real

binding modes form other binding modes
• force fields
• empirical free energy functions
• knowledge based functions

 scoring can be used together with the

conformation search method or only for ranking
the search results

 scoring functions are the most critical issue of

docking

          
 Ligand Preparation in DS

A. Manually

B. Ligand preparation protocol

• Charges are standardized for common groups
• The largest fragment is kept
• Hydrogens are added
• The molecule is represented in Kekule form
• The ionization states may be enumerated
• Tautomers may be generated
• Isomers may be generated
• Duplicates may be removed
• 3D coordinates may be calculated (not at CSC)

          
 Protein Preparation in DS

 Protein health tool

• check your structure
 Protein report and Build and edit protein
• fix your force filed
 Force Field tool
• set up the force filed
• Check the automation level from the protocols from
Edit/Preferences!
 Define and Edit Binding Site tool
• look for cavities
• Define the binding site sphere using a cavity or specific site

          
 Generate ligand
CDOCKER conformations trough
high temperature MD

 CHARMm force filed based docking tool Random (rigid-body)

 Uses soft-core potentials rotation
In the active site
 Confomation serach using simulated
anealing
grid-based simulated
 Grid based energy evaluation
annealing (several
 Force field based scoring cycles)

Full minimization

Output # of refined
ligand poses sorted by
energy

          
 After CDOCKER

 You can
 Rescore the structures using protocol:
• Calculate Binding Energies
 Optimize binding site with the ligand using protocol:
• Ligand Minimization
 Study the results with protocol:
• Analyze Ligand Poses

          
 CDOCKER: Papers to read

Wu G, Robertson DH, Brooks CL 3rd, Vieth M.

Detailed analysis of grid-based molecular docking:
A case study of CDOCKER-A CHARMm-based MD docking algorithm.
J Comput Chem. 2003 Oct;24(13):1549-62.

Erickson JA, Jalaie M, Robertson DH, Lewis RA, Vieth M.

Lessons in molecular recognition:
the effects of ligand and protein flexibility on molecular docking accuracy.
J Med Chem. 2004 Jan 1;47(1):45-55.

Ferrara P, Gohlke H, Price DJ, Klebe G, Brooks CL 3rd.

Assessing scoring functions for protein-ligand interactions.
J Med Chem. 2004 Jun 3;47(12):3032-47.