of H3 loop - Accelrys

Life Science
Webinar Series
Antibody Modeling:
New tools for improved
success
Francisco
Hernandez-Guzman, Ph.D.
November 8, 2007
Sr. Solutions Scientist
fhernandez@accelrys.com

2
Agenda
Antibody Modeling
– Why?
– How?
• Discovery Studio & Pipeline Pilot
• Homology Modeling
– Loop Refinement methods
• Template based
• De Novo based
Conclusion

3
Why Antibodies?
• Critical elements of the immune system as a first line of defense against
foreign entities
• Wide range of uses:
– Clinical: Drug therapies
• In 2003 – over $2.7 billion market

4
The Antibody Structure
• Different types of Ab but same fold Immunoglobulin
• Main focus will be on IgG:
Complimentary
Determining
Regions (CDR’s)

5
The Antibody Structure (cnt’d)
CDR Loops:
Fab
L1
L2 L3
CL
VL
H1
VH
H2 H3
Fc
CH Fv

6
How? Discovery Studio and Pipeline Pilot
• Protocols (Calculations) can be run in Pipeline Pilot or Discovery Studio 2.0

7
WebPort
WebPort
(web
(web
access)
access)
Pipeline Pipeline
Pilot Pilot
(Pro or Lite )
(Pro or Lite )
Chemistry Biology Materials
Accelrys ® Discovery Studio
ISV ISV
Client Client
(e.g.,
(e.g.,
Spotfire)
Spotfire)
Materials
Materials
Studio
Studio
Client
Client
Client Integration Layer
S c i i T e g i i c P l l a t t f f o r r m
Accord
Accord
Discovery
Discovery
Studio
Studio
Client
Client
Tool Integration Layer Data Access Layer
Cmd-Line
Statistics
Reporting ISV
Tools
Discovery
Studio
Accord Accord
Clients Clients
IDBS Oracle ISIS
Databases
Isentris

8
How? (cnt’d)
• Procedure for Antibody Modeling:
1. Sequence based search for homolog
2. Sequence Alignment
3. Building of Homology Model
4. Analysis of Homology Model
5. Identification of CDR loops
6. Refinement of the loops
– Template
– de novo

9
Loop Refinement Method: Template Based
• Model Antibody Loops protocol
– Uses BLAST to find possible templates and alignment to templates
– Based on the alignment, automatically detects hyper-variable loop
regions
– Detects the key signatures if any
– Based on loop similarity and key signature, automatically selects the
best templates for each loop
– Uses MODELER to build new model with “grafted” loop regions
Input structure
Option to construct
homology model
using MODELER

10
Example: Template based refinement of Antibody
hyper-variable loop regions
• Modeling of the 1flr.pdb Fab fragment
– Hypervariable loop detection
– BLAST detected templates
– Key signatures
Key signatures

11
Example: Template based refinement of Antibody
hyper-variable loop regions
• Modeling of the 1flr.pdb Fab
fragment
– Retrieves antibody templates
– BLAST detected templates
– Key signatures
CDR loop Loop length C-alpha RMSD BB-RMSD
L1
L2
L3
H1
H2
H3
12 1.06 1.08
3 0.54 0.50
6 0.56 0.58
7 0.60 0.63
8 0.87 0.94
7 0.55 0.57
“Direct” self hit removed

12
Additional Fab Examples: Template based refinement
• Modeling of: 2ajs.pdb
• Modeling of: 2gcy.pdb
• Modeling of: 2nyy.pdb
CDR loop Loop length C-alpha RMSD BB-RMSD
L1
L2
L3
H1
H2
H3
12 1.36 1.40
3 0.36 0.43
6 0.81 1.24
8 1.30 1.62
5 2.16 2.00
10 4.08 3.84
CDR loop Loop length C-alpha RMSD BB-RMSD
L1
L2
L3
H1
H2
H3
11 1.00 1.03
3 0.91 0.87
6 1.04 1.33
7 0.54 0.57
6 0.57 0.53
9 1.67 1.92
CDR loop Loop length C-alpha RMSD BB-RMSD
L1
L2
L3
H1
H2
H3
11 0.93 0.94
3 0.65 0.64
6 0.57 0.63
7 0.73 0.73
6 0.84 0.82
9 2.64 2.68

13
Summary:
• The “Model Antibody Loops” protocol successfully automates the
tedious task of antibody loop grafting
• Automated hyper-variable loop detection helps to quickly identify
the loops of interest
• Model building is incorporated seamlessly to the process using
MODELER

14
Loop Refinement Method: de novo
• Full loop reconstruction without prior knowledge
• Use of CHIROTOR and LOOPER algorithms
– CHIROTOR
• CHARMm based algorithm
• Automatically rebuild side chains de novo
– LOOPER
• CHARMm based algorithm
• Automatically reconstruct loop regions de novo

15
Side-Chain Refinement using ChiRotor
• ChiRotor side-chain
refinement:
– Performs systematic
search side-chain
conformations
– Minimizes
conformations using
CHARMm
– Selects the best
conformation based on
CHARMm energy
• Fast abintioapproach
not dependent on the
starting structure
– Rebuilds the sidechains
Animated figure

16
ChiRotor
Methodology
Protein
3D Structure
Select a Set of n
Residues
For Refinement
Remove all side
chain atoms of
selected residues
Start loop for i from 1 to n
Choose Residue i
Sample side chain
conformations of
residue i varying χ 1
Energy Minimize side
chain atoms of residue
i in CHARMm
Save 2 Best
Conformations for
residue i
End loop for i
Output: 2n partial
structures
Construct complete
structure using lowest
energy conformer of each
residue.
Energy minimize all
selected side chains
Start loop for i from 1 to n
Replace side chain
conformation i with the 2 nd
conformer and energy
minimize
Accept the structure if energy
is lower.
End loop for i
Output: 1 Lowest Energy
structure

17
Loop Refinement using Looper
• The Looper loop
refinement algorithm
– Systematically
searches for backbone
conformation
– Uses CHARMm
minimization
– Ranks the
conformation using
CHARMm energy,
including solvation
energy term
• Fast, ab intio approach
not dependent on the
starting structure
– Rebuilds the loop
Loop Refinement protocol and results in Discovery Studio

18
Looper Methodology
• Looper first constructs and optimizes the loop
backbone
– Systematic search of loop conformation by
sampling a minimum set of backbone dihedral
angles φ and ψ
– The loop is divided into two halves and each half
is constructed independently from the end of the
loop and then combined without the side-chain
atoms
– The loops are minimized by CHARMm and ranked
by CHARMm energy
Continued…

19
Looper Methodology
• Next, Looper constructs the loop side chains
and optimizes the loop
– Construct the side-chains using the ChiRotor
and the loop conformations are ranked by
CHARMm energies
• Finally, Looper re-ranks the conformations
– CHARMm energy minimizations in the first two
steps are done without including the solvation
energy term
– In this step, each top ranking loop
conformation is re-scored by adding the
solvation energy term calculated using
Generalized Born approximation (in CHARMm)

20
Example: Homology Modeling of an Antibody fragment and
de novo reconstruction of H3 loop
• Modeling of the 2aab Fab fragment
– Clinically relevant melanoma antigen system
– 25 models built, high quality
– 1mf2 used as template
– 89.4% id, 92.8% sim
8 residue H3 loop modeled!!!
Fab
RMSD* (seq): 2.46, 434 residues
RMSD* (struct): 1.94 279 residues
Fv
RSMD* (seq): 1.27, 232 residues |0.90 224 residues * †
RMSD* (struct): 0.83, 224 residues | 0.76, 220 residues* †
* backbone atom rmsd † no H3 loop included

21
MODELER
• Best Homology model:
Fv RMSD (seq) Fv RMSD (seq) MODELER
1.27 (232 resid)
Green = 2aab cryst.
Red = 1mf2 cryst.
Orange = 2aab H. Mod.
0.90 (224 resid)
H3 loop not incl.
Fv RMSD (seq)
1.18 (232 resid)
MODELER
Fv RMSD (seq)
0.90 (224 resid)
H3 loop not incl.
H3 loop RMSD
4.11 (8 resid)

22
Loop refinement
Mutate H3-loop to Ala
Side Chain refinement of residues within
8Å radius (ChiRotor*)
Mutate H3-loop residues back
Loop Refinement (LOOPER**) of H3 loop
Case:
Pre-Ref.
bb RMSD
Post-Ref.
bb RMSD
No Neighbor
opt.
4.11Å
3.81Å sol #1
3.02Å sol#3
Neighbor opt.
4.11Å
1.79Å sol #1 (3)
1.54Å sol #21 (25)
*Spassov, V. et al. Protein Science. 16:1-13 2007
** Spassov, V. et al. (manuscript)
2aab crystal
struct.
1.30Å
X-ray structure
post-refined
pre-refined
side view
top view

23
Summary:
• Success of LOOPER strongly correlated to proper positioning of
neighboring side-chains around the loop
• ChiRotor successfully refined neighboring side chains for better loop
refinement
• Need to generalize our method with more test cases

24
Conclusion
• Rules based recognition of hyper-variable loop regions can be used to
quickly annotate and properly align sequences for template based
modeling
• Template based methods can model hyper-variable loop regions with good
accuracy
• De novo loop refinement methods are not bound by template bias and
also show a high degree of success for rebuilding the highly flexible
hyper-variable loops such as the H3 loop of the 2aab antibody fragment
• Success in our de novo loop reconstruction calculation was directly
correlated with the proper positioning of the neighboring side chains
• Two methods represent the full solution for Antibody loop modeling

25
Acknowledgements
• Velin Spassov
• Zeljko Dzakula
• Yi-Shiou Chen
• Lisa Yan
• Paul Flook
• Shikha Varma-O’Brien

26
References:
• Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
Swaminathan, S.; Karplus, M. CHARMM: A program for
macromolecular energy minimization and dynamics calculations.
1983, J. Comp. Chem. 4: 187-217
• Morea V., Lesk A. and Tramontano A. Antibody Modeling:
Implications for Engineering and Design. 2000, METHODS 20: 269 –
279
• Sali A. & Blundell T.L. Comparative protein modelling by
satisfaction of spatial restraints. 1993, J. Mol. Biol. 234: 779-815
• Spassov V., et al. The dominant role of side-chain backbone
interactions in structural realization of amino acid code. ChiRotor:
A side-chain prediction algorithm based on side-chain backbone
interations. 2007, Protein Science. 16: 494-506