Prediction of pKa shifts in proteins using a discrete rotamer search and the Rosetta energy function

1

Prediction of pKa shifts in proteins using a
discrete rotamer search and the Rosetta energy
function
Ryan M Harrison, Jeffrey J Gray
Baltimore Polytechnic Institute Johns Hopkins
University, Department of Chemical Biomolecular
Engineering
2
pH has profound effects on proteins

• Conformational Change
• Catalytic activity
• Binding affinity
• Stability

Influenza Hemagglutinin protein
Red pH-sensitive region of hemagglutinin
Harrison RM 2005
3
Rosetta Algorithm
Protein Folding
Harrison RM 2005
4
Objective
Improve computational protein structure
predictions by describing how proteins react to
different pH environments

• Develop and implement pH-sensitive modeling in
  Rosetta
• Predict pKa shifts in several model proteins
• Model pH-sensitive docking and folding
• Design a protein with pH-sensitive activity

Harrison RM 2005
5
Why model pH in Rosetta?

• More accurate predictions
• Enhanced description of protein energy landscape
  
• More physically relevant protein electrostatics,
  especially __buried charges

• Extended Capabilities
• Predict pH-sensitive conformational changes
• Sidechain, Backbone, Rigid Body (?)
• Predict docking and folding pH-optimums
• Design novel pH-sensitive motifs and functions

Harrison RM 2005
6
Develop the framework
Improve computational protein structure
predictions by describing how proteins react to
different pH environments

• Develop and implement pH-sensitive modeling in
  Rosetta
• Predict pKa shifts in several model proteins
• Model pH-sensitive docking and folding
• Design a protein with pH-sensitive activity

Harrison RM 2005
7
pKa shifts
pH titration (Idealized)
pKa shift
IpKa
pKa
pKa The pH at which an amino acid equally
occupies its protononated and deprotonated states
Harrison RM 2005
8
Methodology
Harrison RM 2005
9
Procedure

• Allow Rosetta to dynamically select most
  favorable amino acid protonation state
• Introduce an energy function for protonation
• 2. Allow Rosetta to sample alternate protonation
  states
• 3. Modify amino acid parameters for each state

Harrison RM 2005
10
Rosetta Score Functions
Harrison RM 2005
11
Predict pKa shifts
Improve computational protein structure
predictions by describing how proteins react to
different pH environments

• Develop and implement pH-sensitive modeling in
  Rosetta
• Predict pKa shifts in several model proteins
• Model pH-sensitive docking and folding
• Design a protein with pH-sensitive activity (?)

Harrison RM 2005
12
Model Systems
Turkey Ovomucoid Inhibitor (OMTKY3)
Ribonuclease A (RNaseA)
Harrison RM 2005
13
Ribonuclease A
pKa shift
Harrison RM 2005
14
Turkey Ovomucoid Inhibitor
Rosetta predicts pKa shifts with 0.77 root mean
squared (rms) accuracy
Red Rosetta Prediction, Green Experimental,
Gray IpKa (Null Value)
Harrison RM 2005
15
Turkey Ovomucoid Inhibitor
LYS29
ASP27
CPK Prediction, Green Experimental Rosetta
under shifted pKas
Harrison RM 2005
16
Ribonuclease A
Model rms eprotein IpKa 0.95 Rosetta 0.62 er SCC
E 2.69 4 MCCE 0.99 4 MCCE 0.66 8 MCCE 0.44 20
Rosetta predicts pKa shifts with 0.62 rms accuracy
Red Rosetta Prediction, Green Experimental,
Gray IpKa (Null Value)
Harrison RM 2005
17
Ribonuclease A
HIS12
CPK Prediction, Green Experimental Rosetta
predicted pKa precisely
Harrison RM 2005
18
Ribonuclease A
Low pH
High pH
Predicted pKa 3.5 Experiment 3.5 IpKa
4.0
ASP 83
ASP 121
HIS 119
Harrison RM 2005
19
Conclusions
Rosetta can now estimate the local effects of pH
(i.e. pKa shifts) in small globular proteins
Developed an approach to model pH Accounted for
significant pKa shifts using only side-chain
movement Extended the modeling capabilities of
Rosetta Increased the overall accuracy of
Rosetta(?)
Harrison RM 2005
20
Work in Progress

• Optimization and calibration on a set of over
  200 experimentally determined pKa shifts from 15
  proteins
• pH-sensitive Docking and Folding
• Scientific and performance benchmark on 55 pKas
  from staphylococcal nuclease mutants (in
  collaboration with Garcia-Moreno lab)

Staph. Nuclease at pH 7.2
?-helical nano-gel
Harrison RM 2005
21
pH-sensitive docking
Improve computational protein structure
predictions by describing how proteins react to
different pH environments

• Develop and implement pH-sensitive modeling in
  Rosetta
• Predict pKa shifts in several model proteins
• Model pH-sensitive docking and folding in
  several model proteins
• Design a protein with pH-sensitive activity (?)
  

Harrison RM 2005
22
Acknowledgements
National Institutes of Health National Institute
of General Medical Sciences Gray Lab Dr. Jeffrey
J. Gray Harden Lab Dr. James L.
Harden Baltimore Polytechnic Institute The
Ingenuity Project Ms. Charlotte V. Saylor Robert
M Harrison Sharon A Harrison
Harrison RM 2005
23
(No Transcript)
24
Harrison RM 2005
25
What could proteins do for you?
Drug Design
Imagine targeted treatments for devastating
diseases
Blue antibody, Red prediction, Green
experimental Antibody binding to ovine prion.
Figure from M Daily, Pymol
26
Rosetta Score Functions Electrostatics

• Electrostatics require electron density
  parameters
• Predictions were made using both a Generalized
  Born (GB) and Coulombic electrostatic model.
• GB electrostatics are more accurate than
  Coulombic electrostatics, but also more
  computationally expensive

Harrison RM 2005
27
Rosetta Procedural Detail
Rosetta Flowchart
Low Resolution _1. Rigid Body Move
_ _2. Monte Carlo
Minimization High Resolution _1.
Sample all side chain positions in ___Dunbrack
rotamer set 2.
Sample alternate protonation ___state rotamers
_ _3. Monte
Carlo Minimization Post-Processing
1. External Scripts to
determine side ___chain pKa values
Start Position
Low Resolution Monte Carlo
High-Resolution Refinement
10n
Post-Processing
Predictions
Modified to introduce pH-sensitive side chain
modeling or pKa predictions in Rosetta
Harrison RM 2005