Nucleic Acids

1
Nucleic Acids
Chuck Duarte, Jim Havranek
2
Interface Mode
Tanja Kortemme
3
INTERFACE MODE -interface
Binding energy (-interface -s ltpdbnamegt
-ddg_bind_only) DGbind DG(complex) -
DG(partnerA) - DG(partnerB)
CONTRIBUTIONS TO THE BINDING ENERGY FOR THE WT
COMPLEX ---------------------------------------
--------------------------------------------------
----------------------------------------
Eatr Erep Esol Eaa Edun Eintra
Ehbnd Epair Eref Eh2o Eh2ohb Eh2osol
Eres pdb name -------------------------------
--------------------------------------------------
------------------------------------------------ D
G_BIND -13.8 0.3 6.7 0.0 0.0
0.0 -1.1 0.0 0.0 0.0 0.0
0.0 -9.4 1be9 --------------------------
--------------------------------------------------
--------------------------------------------------
-----
4
INTERFACE MODE -interface
Changes in binding energy upon mutation
(-mutlist) DDGbind DGbind(mutant) -
DGbind(WT)
BINDING
ENERGY FOR MUTANT COMPLEXES
CONTRIBUTIONS TO THE BINDING
ENERGY FOR THE MUT COMPLEXES Eatr
Erep Esol Eaa Edun Eintra Ehbnd
Epair Eref Eh2o Eh2ohb Eh2osol Eres
ROS PDB CH WT MUT BIND_CON -13.8 0.3
6.7 0.0 0.0 0.0 -1.1 0.0
0.0 0.0 0.0 0.0 -9.4
WT BIND_CON -13.1 0.3 6.6 0.0
0.0 0.0 -1.1 0.0 0.0 0.0
0.0 0.0 -8.8 27 327 A I
A BIND_CON -13.3 0.3 6.0 0.0
0.0 0.0 -0.8 0.0 0.0 0.0
0.0 0.0 -9.2 118 7 B T
A BIND_CON -12.1 0.3 6.8 0.0
0.0 0.0 -1.1 0.0 0.0 0.0
0.0 0.0 -7.6 120 9 B V
A
CHANGES IN BINDING ENERGY FOR MUTANT
COMPLEXES MUT-WT
CHANGES IN CONTRIBUTIONS
TO THE BINDING ENERGY FOR THE MUT COMPLEXES
Eatr Erep Esol Eaa Edun
Eintra Ehbnd Epair Eref Eh2o Eh2ohb
Eh2osol Eres ROS PDB CH WT MUT Nei DDG_BIND
0.7 0.0 -0.1 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
0.6 27 327 A I A 27 DDG_BIND 0.5
0.0 -0.7 0.0 0.0 0.0 0.3
0.0 0.0 0.0 0.0 0.0 0.2 118
7 B T A 19 DDG_BIND 1.7 0.0
0.1 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 1.8 120 9 B
V A 24
5
INTERFACE MODE - PARAMETERIZATION

• SIDE-CHAIN MODELING
• structural relaxation upon mutation (repack
  neighbors)
• repacking of partners
• no significant improvement for ala scanning

• ENERGY FUNCTION
• int parameterized on DDGbind values in
  interfaces (Lin)
• sc-sc HB term environment-dependent (linear -
  Lin)
• pack parameterized on aa identities
  (packer weights )
• HB now default env-dep
• default should be pack for pack_rotamers and
  int for score
• currently int for both
• -Wint_only int for pack and score
• -Wpack_only pack for pack and score

6
INTERFACE MODE - WEIGHTS
Ala!
Ala!
gt needs parameterization/benchmarking
7
INTERFACE MODE - ENERGY FUNCTION
restrictive sc-sc HB (lt 2.3 Å) works
better specificity prediction
8
Alter Specificity
Brian Kuhlman
9

• Alter_spec mode version of second-site
  suppressor strategy recently used by Tanja to
  create orthogonal binding interactions.
• Goal redesign two interacting proteins so that
• DGMut-Mut DGWT-WT ltlt DGWT-Mut DGMut-WT
• Protocol
• Loop through all possible point mutants at the
  interface and identify mutations that weaken
  binding. Neighboring residues are allowed to
  repack to accommodate the point mutations.
• For each destabilizing mutation redesign the
  surrounding residues to better accommodate the
  mutation. Output a file (mutlist) that contains
  all the point mutations and the compensating
  mutations.
• Calculate the binding energies of WT-WT, WT-Mut,
  Mut-WT and Mut-Mut for each redesign using
  interface mode with the alter_spec flag. The
  results are sorted so the most promising
  redesigns are at the top.

Deanne Sammond
10
Sample command lines requires two steps
1) rosetta.gcc -s 1c4z.pdb -design -alter_spec
-pmut pmut63 -alter_spec_mutlist mut63c -fix
fix63n -alter_spec_mutlist file name
output file with list of point mutants and
redesigns -fix file name input file
specifying residues that are arent allowed to
change identity -pmut file name input file
specifying which residues to try point mutants at
(the default is all interface residues) 2)
rosetta.gcc -s 1c4z.pdb -interface
-repack_neighbors -alter_spec_format -mutlist
mut63c -intout E63c -mutlist file name
input file generated by the previous step
11
Output
point_mut_number 4 structure_number 10 ,
11 , 12 Eatr Erep Esol
Eaa Edun Eres Gmut-Gwt Partner1 5.4
-2.4 0.0 -0.7 -2.2 -1.1 Gmut-Gwt
Partner2 1.4 -0.1 -0.2 -0.4 0.3
-1.0 ddGbind MUTMUT -4.8 -4.7 2.8
0.0 4.0 -5.3 ddGbind MUTWT -5.0
4.7 1.3 0.0 2.0 4.9 ddGbind
WTMUT 1.4 -0.4 2.5 0.0 0.7
1.6 Mutations Partner 1 S638A I655W F690Y Y694A
Mutations Partner 2 F63E point_mut_number
3 structure_number 7 , 8 , 9
Eatr Erep Esol Eaa Edun
Eres Gmut-Gwt Partner1 1.9 -2.0 3.9
-0.6 1.1 1.9 Gmut-Gwt Partner2 1.1
-0.1 0.3 -0.4 -0.1 -0.9 ddGbind
MUTMUT -1.3 -6.4 2.2 0.0 0.9
-9.0 ddGbind MUTWT -4.7 -3.2 1.6
0.0 0.5 -3.6 ddGbind WTMUT 1.5
-0.6 2.0 0.0 -0.5 0.9 Mutations
Partner 1 S638A L642R I655W Y694A Mutations
Partner 2 F63D
12

• Observations
• often redesigns residues that dont interact
  directly with the point mutation
• the protocol does not explicitly optimize
  affinity so sometimes the redesign just enhances
  interactions within each chain.
• having trouble finding designs that destabilize
  both WT-Mut and Mut-WT

13
Symmetry Mode
Ora Furman
14
Symmetric docking mode
15
INTERFACE MODE - ENERGY FUNCTION
restrictive sc-sc HB (lt 2.3 Å) works better
monomeric proteins
more stringent hydrogen bonding
16
Symmetric docking of Homo-multimers basics
Regular docking docking frame is varied monomer
1 is fixed
Symmetry docking docking frame is fixed monomer
1 is varied

• Create symmetric multimer
• Rotate monomer 1
• other monomers (contact point on x-axis
  angle360o/n)
• Optimize distance T

17
Symmetric docking outline
read in monomer1, center on origin of
coordinates. Read in docking_T_size. OR read in
multimer, define docking_T_size, the translation
and the rotation axes
18
Symmetric docking command line options

• -symmetry
• evokes symmetry mode sets docking_symmetry T
• -n_monomers
• number of monomers in complex (default 2
  current max3)
• -init_T_size
• initial distance of center of monomer from
  contact point center of multimer complex
  defines initial docking_T_size
• -multimer
• indicates that multimer is read in. sets
  dock_sym_multimer_start T
• Used to derive symmetry frame.

19
Symmetric docking notes

• Currently assumes C symmetries (planar)
• Interface residues are evaluated between monomers
  1-2 only and assumed to be the same for all
  monomer x - x1 pairs
• Backbones are symmetric, side chains are
  monomer-specific
• Side chain conformations are stored and optimized
  for each monomer separately
• Elongation
• The degree of elongation (i.e. planarity) can be
  monitored (external script) to enrich for wanted
  conformations
• RMSD calculation
• Since monomer1 changes (instead of the docking
  frame), the structure has first to be moved so
  that monomer1 is superimposed onto native
  monomer1

20
(No Transcript)
21
Flexible Docking Calculations
22
Backbone Conformational ChangeCAPRI T01 HPr
HPr Kinase (Round 1, Sep 2001)
Terminal helix swings upon docking, nuzzling HPr
in a pocket
No energy funnel for binding the unbound
components
23
Torsion Angle Perturbation
C-terminal helix a4
Torsion angle movement in residues 290-292 would
allow the correct conformation to be observed.
Res 290-292
Kinase I
24
Low-Resolution Search With Flexible Backbone
Initial helix perturb
Initial rigid body perturb
Rigid-body move
Helix Minimization
Helix MC perturb
Monte Carlo Accept?
Small helix perturb
10x50
Low resolution output decoy
25
Flexible Docking Results With torsion angle
perturbations and explicit minimizations
score
HPr rmsd
score
18/36 contacts, translation 1.8Å, rotation 18º
helix rmsd
26
Loop Flexibility

• Currently exploring ways of moving loops during
  protein-protein docking to simulate an induced
  fit binding mechanism

Rohl, CA et al 2004 to appear
27
Docking Algorithm Overview
Random Start Position
Build/Optimize Loops
Low-Resolution Monte Carlo Search
High-Resolution Refinement
105
Clustering
Predictions
28
Implementation

• Call loop mode to build/optimize loops before
  low-res docking search
• (Future work integrate loop moves during low-res
  docking search, alternating with rigid-body moves)

29
Surfaces
Dave Masica
30
Redesigning Proteins for Optimal Interactions
with Inorganic Crystals
David Masica from Dr. Jeff Grays Lab
31
Generation of a Crystalline solid from the Unit
Cell
5.630 Å
Cl
Na1
32
Overhead and Side Views of Pre-docked Ferritin
and NaCl Crystal
33
Current Working Flowchart
Start Position
500
Best Scoring Decoy
Define Contact Residues and Design (Design
Ligand)
Best Scoring Decoy
10
Post-Process (Based on score and homology)
34
Ferritin NaCl Complex after many Dock and Design
Attempts
Res Original Mutant 4 ILE
ASP 53 GLU ILE 60 GLU
ARG 64 ARG LEU 67 LYS
ILE 71 GLN LEU 124 HIS
ASP 127 ASP ILE 128 PHE
TYR 130 GLU ILE 131 SER
LEU 132 HIS TYR 135 ASP
VAL 136 GLU VAL 168 ARG
LEU 169 LEU THR 170 THR
VAL 172 LYS ALA 173 HIS
ARG 174 ASP LEU
88.5 identical To native ferritin
Magenta Mutated Residues
35
Questions and Challenges to Come

• Formation of covalent bonds between cysteines and
  inorganic hetero atoms
• Is a new approach necessary to define heavy atom
  types in Rosetta?
• Modeling inorganic crystal interactions with
  water
• Modeling many proteins interacting with a single
  crystal simultaneously
• Benchmark?
• Modeling the loose electrons in metal compounds

36
Jumping Code
Phil Bradley
37
Docking with more flexibility

• Chu Wang
• 2nd Rosetta Meeting
• August, 2004

38
Whats New?

• Rosetta Docking
• Include unbound native rotamers.
• Rotamer trials minimization.
• Refolding with native bond length/angle.
• Docking with BB/SC minimization.
• Super predictions in CAPRI.
• Chu Wang
• A married man.
• A PhD candidate.

39
Side-chain modeling in docking too left or too
right?

• All side-chains used to be removed and repacked
  with rotamers.
• Use rotamer trials minimization to sample
  off-rotamer space.
• Include native unbound rotamers to increase
  side-chain conservatism.
• Combines them to improve docking performance.

40
Repacking of interface side-chains
41
Improving docking perturbations
42
Refold with native bonds

• The minimized native structure without
  idealization first usually blows up when template
  bond parameters are used.
• But for docking and homology modeling, we do have
  input bond information and they should be used as
  much as possible.
• With the new_refold algorithm, native bond
  information is stored and recalled for folding
  later.

43
(No Transcript)
44
A good reason to explore flexible-backbone
docking
45
Rigid-body Docking does not work all the time

46
In order to fit , they have to BEND!!!
Therefore, we know the importance of backbone
flexibility
47
1ACB Chymotrypsin / Eglin C
Chain E I
Chain A B
48
Backbone minimization of this loop
49
How best can we do?
50
How well did we do?
51
Lessons and Plans

• Rigid-body movement seems to be able to
  compensate this loop deviation between bound and
  unbound.
• Backbone minimization works only when the
  rigid-body position is right, which is hardly to
  be sampled.
• Collect more test cases to apply BB/SC
  minimization with limited RG search.

52
Incorporating Side Chain Entropy into an Energy
Function for Protein Design

• Jenny Hu Brian Kuhlman

53
Rationale for calculating side chain entropy

• Currently it is being ignored
• Expect that it will change the relative
  favorability for burying different amino acids
• A MET will lose more entropy when buried than a
  VAL will

54
How to calculate entropy
Repack the protein, while repacking, keep track
of how often each rotamer is observed and
calculate the average energy for each residue
R gas constant Pi the probability of the side
chain being in rotamer class i
55
Result of entropy calculation (simulation is
based on 2661 proteins, 622,366 residues)
56
Design with explicit side chain entropy

• Pick a random sequence
• Repack the sequence to calculate ltUgt and TS
• Force an amino acid mutation and repack the new
  sequence to determine the new ltUgt and TS
• Evaluate move by comparing new ltUgt - TS with old
  ltUgt - TS
• Goto (3)

57
Modified 2-layer approach
58
Thank you !
59
1et9
LYS
SER
60
(No Transcript)
61
Computational design of artificial catalytic
protein

• Lin Jiang
• David Bakers lab
• 08/09/04

62
Outline

• Inversive Rotamer Tree approach
• (1) Tansition State (TS)
• (2) Building inversive rotamer tree around TS
• (3) Generating backbone coordinates for all no
  clashing sidechain rotamer combination
• (4) Identify sites in a set of possible protein
  scaffolds matching each of the backbond
  coordinates in step (2)

63
Outline

• Inversive Rotamer Tree approach
• (5) for each identified site, computationally
  designing the amino acid packing around TS while
  retaining catalytice functional group geometry
• (6) Optimizing placment of TS model in binding
  site by rigid boby docking together with protein
  design.
• (7) select lowest energy design with best
  catalytic site geometry. QM evaluation of
  designed site and refine it

64
1.
3.
1. TS Ligand Coordinate
Protein Coordinate
2. Inversive Rotamer Ensemble
3. Polyalanine Chain
2.
4. Placed Vrotamer Ensemble
4.
5. Ranked Vrotamer Ensemble
5.
6. Pocket Sidechain Design (PSD)
6.
65
Recover the active site of native enzyme TIM
66
More match TIM
Dwyer et al, Science. 2004304(5679)1967-71.
67
Recovery of native aldolase
68
Design of Unnatural Aldolase
69
Acknowledgement
David Baker (PI) Jens Meiler Gong
Cheng Brian Kuhlman Jeffrey J.
Gray Tanja Kortemme Alexandre
Zanghellini Baker lab members BMSD program and
Dept. of Biochemistry
70
(No Transcript)
71
Disulfides
Bill Schief
72
Pi-pi and aliphatic orientation-dependent scoring
Kira Misura
73
Plane orientation score Pi-pi (F/Y/W), cation-pi
(F/Y/W,H/R), proline, hydrophobic

• Cation-pi partial negative charge on benzene
  ring, positive charge on lys/arg
• Pi-pi quadrupole moment, strongest in T-shaped
  or offset stacked conformation- electrostatics
  dominate?
• Proline?
• Hydrophobic- capturing something about packing
  that lj is missing? Minimization of this term
  might improve rotamer selection, packing.

74

• Get statistics from PDB for native proteins
• Xtal structures, gt2.5A resolution
• Contact definition at least 2 heavy atoms within
  4.2A.
• Decoy set 70 proteins, 500 decoys each

75
Define distances and orientation between planes
in sidechains
cross2
center_center
q
cross1
cross2
cross1
vertical
horizontal
76
FF - offset stacked
FF T-shaped
WP, stacked
WH, stacked
77
Compare natives to decoys500 decoys from 70
proteins,centroid relax fullatom relax
FF decoy
FR decoy
FR native
FF native
vertical
horizontal
Blue 0-30 deg. Green30-60 deg. Red60-90
deg.
78
FH decoy
FP decoy
FP native
FH native
vertical
horizontal
Blue 0-30 deg Green30-60 deg Red60-90 deg
79
Class R RF,RY,RW,RR
vertical
horizontal
80
vertical
horizontal
Class P FP,YP,RP,HP,WP
81
Native ranks ( of proteins where native receives
score in the top 15)
Energy gap change between native and decoys-
comparison with LJ energy and LJ energy plane
score energy
82
Generalized BornElectrostatics
Jim Havranek
83
Generalized Born Electrostatics
Implicit water continuum dielectric We want an
environment-dependent ES model, but the irregular
geometry of the protein-solvent interface makes
this hard. GB is one way to encapsulate the
environment of an atom.
84
Generalized Born Radii are measures of burial
Born radius for ion
GB model associate with each atom a concentric
sphere that gives an equivalent degree of burial
as in the complex protein geometry. (Invert the
equation above).
85
Charged interactions
Coulombic
Polarization
where
Limiting behavior at d0, reverts to Born
equation as d approaches ?, reverts to
Coulombs law
86
GB in Rosetta
Using the method of Onufriev, Bashford, and Case
to get generalized Born radii. Not directly
pairwise additive - one pairwise pass gets you
radii, another gets you energies. Design mode
positions with uncertain conformations/identities
are given a placeholder sphere for radii
calculation. Code doesnt execute if Wgb_elec
0.0
87
Sequence recovery with GB electrostatics
Good Asp, Glu, Lys, Arg Split Asn good, Gln
neutral Bad His Overall 1.5 better
88
Sequence recovery in protein-DNA interfaces
Good Glu, Lys, Arg Bad His, Trp Overall 3.0
better
89
Summary
GB is working in Rosetta improves sequence
recovery, decoy discrimination slower than other
terms Under construction derivatives weights
with ligands Should be useful for enzyme design
/ pKa prediction
90
FULL ATOM SCORE FUNCTION
SCORE 12 (current fa scorefxn) For protein type
xx hb_lrbb 1.0 (Long range HB) hb_srbb
0.5 (Short range HB) For protein type a
(alpha) hb_lrbb 1.0 hb_srbb 0.5 For all
others (b and ab) hb_lrbb 2.0 hb_srbb 0.5
hb_sc 1.0 (Sidechain HB) fa_atr 1.0 (LJ
attractive) fa_rep 1.0 (LJ repulsive) fa_dun 1
.0 (Dunbrack term) fa_pair 1.0 (Pair
term) fa_sol 1.0 (LK Solvation
term) fa_prob 0.5 (akin to Rama
term) rama 0.2 (Ramachandran term)
SCORE 13 (NEW fa scorefxn) For protein type xx
hb_lrbb 0.7 (Long range HB) hb_srbb 0.1
(Short range HB) hb_sc 0.8 (Side chain HB) For
protein type a (alpha) hb_lrbb
0.1 hb_srbb 0.4 hb_sc 2.5 For protein type
b (beta) hb_lrbb 1.1 hb_srbb 0.1 hb_sc 0.
3 For protein type ab (mixed alpha-beta) hb_lrbb
0.7 hb_srbb 0.1 hb_sc 0.7 fa_atr 1.0
(LJ attractive) fa_rep 1.0 (LJ
repulsive) fa_dun 0.5 (Dunbrack
term) fa_prob 2.0 (akin to Rama
term) fa_gb_elec 1.3 (Gen Born Electrostatics
term) sasa 0.05 (Surface area)
New solvation term
91
CENTROID SCORE FUNCTION
SCORE 6 (previous centroid scorefxn) For protein
type xx hb_lrbb 0.5 (Long range HB) hb_srbb
0.25 (Short range HB) For protein type a
(alpha) hb_lrbb 0.5 hb_srbb 0.25 For all
others (b and ab) hb_lrbb 0.25 hb_srbb 1.
0 vdw 1.0 (VDW attractive) env 1.0
(Solvation term) pair 1.0 (Repulsion
term) cb 1.0 (C beta term) sheet 0.5 (Beta
sheet term) ss 0.5 (sheet-sheet term) hs 1.0
(helix-sheet term) rsigma 0.5 rg 1.0 (Radius
of gyration) rama 0.1 (Ramachandran term)
SCORE 6 7 (NEW centroid scorefxn) For protein
type xx hb_lrbb 1.3 (Long range HB) hb_srbb
0.4 (Short range HB) For protein type a
(alpha) hb_lrbb 1.3 hb_srbb 0.7 For all
others (b and ab) hb_lrbb 1.3 hb_srbb 0.1
vdw 1.0 env 1.0 pair 0.4 cb 1.5
sheet 0.0 ss 0.3 hs 0.0 rsigma 0.0
rg 1.0 (Radius of gyration) rama 0.1
(Ramachandran term)
92
Mixed alpha-beta
93
All Beta
94
All alpha
95
Constraint Handling
Carol Rohl
96
Restraints

• Types
• NOEs Distance restraints on atom pairs
• RDCs Orientational restraints on atom pairs
• Checked for automatically in all Rosetta modes.
• Default filenames 1pdb_.cst, 1pdb_.dpl
• Command line flags -cst -dpl change the
  default extensions
• Mode-specific behavior
• Ab initio folding uses modified protocols in the
  presences of restraints.
• refine mode is a centroid-based refinement
  primarily for optimizing distance restraints.

97
Distance Restraints Format

• Reference NMRformats.README in the
  rosetta_documentation archive
• Specify residue number and atom name for two
  atoms, and lower and upper distance bounds.
• Any IUPAC atom name acceptable and/or CEN for
  centroid-based constraints.
• Degenerate hydrogens
• H atoms with a shared heavy atom (methyl,
  methylene, amino)
• Hydrogen number replaced with .

98
Evaluation of Restraints

• Side-chain constraints evaluated only in
  full-atom scoring schemes.
• If NO centroid constraints are provided
• weak centroid constraints are automatically
  generated and evaluated in centroid scoring
  schemes
• 15Å upper bound
• Parameters (constraints_ns.cc)
• MAX_CONSTRAINTS max of distance restraints
• DEGENERATE_PAD upper bound pad for degenerate
  restraints
• GLOBAL_PAD upper bound pad added to all
  restraints
• WARNING PACKER IS RESTRAINT-UNAWARE!!

99
Potentials

• Four scoring schemes
• 1. Linear penalty for exceeding upper bound,
  truncated at 10Å
• 2. Linear penalty without truncation
• 3. Quadratic for small violations, linear for
  large (CNS potential)
• 4. Quadratic penalty for exceeding upper bound
• Select with score_set_cst_mode(int mode)
• Lower bound matters only for CNS-type potential
  (mode3).

100
Extras

• main_frag_cst_trial
• Fragment insertions with prescreening for
  fragments that will improve the constraint score
• constraint stage
• Selective evaluation of restraint subsets
  according to sequence separation.
• Pairs with sequence separation lt stage are
  evaluated.
• Set with score_set_cst_stage(int stage)
• Used primarily in de novo folding.