Transcript The Protein Folding Problem When will it be solved?

The Protein Folding Problem
When will it be solved?
Dill et al, 2007
What is protein folding?
Why is it a problem?
What are some approaches to understanding it?
How far have we come?
What does the future hold?
Protein folding is how an amino
acid sequence (a polypeptide)
folds into its native structure.
A native structure is the functional form of a protein.
The overall protein folding
problem is to understand the
relationship between amino-acid
sequence and protein structure.
The protein design problem is to
synthesize a stable amino-acid
sequence towards a target
conformation.
Three ‘Easy’ Pieces
• Understanding how inter-atomic forces
contribute to a protein’s native structure.
What is the driving force behind protein folding? Does
understanding the problem have other levels of applicability?
• Prediction of native structure from a given
amino acid sequence.
Can we input a polypeptide sequence and output the ‘correct’
protein? How accurate would this simulation be?
• The kinematic question of folding speed
Just how do proteins fold so fast? Can we attain this speed
and accuracy with synthetic, de novo proteins?
Piece One
Understanding Folding
• Before mid-1980s, overall folding process was
seen as sum of small, local interactions.
Like hydrogen-bonds, van der Waal’s interactions, ion pairs.
• Statistical mechanical modeling after mid-80s
changed this view.
• Big component is reducing exposed hydrophobic sidechains. I.e.,
non-local interactions are the ‘driving force’.
• Folding process is distributed locally and non-locally
• Free Energy Equation
• Effects of cytosol cannot be ignored.
• Composition (solvent, other macromolecules, pH)
• Temperature
Hydrophobic Interactions
What is the applicability?
• Design of foldamers
Synthetic molecules which mimic folding ability of
proteins (e.g. peptoids in pharmaceutics)
•
•
•
•
•
Design of lung surfactant replacements
Cytomegalovirus inhibitors
Antimicrobials
siRNA delivery agents
Synthetic proteins from “broadened
alphabets”
Piece Two
Ab initio Structure Prediction
Bioinformatics-based approach
• Long-standing goal.
Amino-acid sequence in → 3-D Native Structure out
• Makes drug discovery faster.
Simulate drug interactions without costly studies.
• Makes it cheaper.
Replace experimental structural determination with accurate
computer simulation.
CASP
• Critical Assessment of Techniques for
Structure Prediction (Moult, 1994)
• Social experiment
• Prediction of native state from amino-acid
sequence alone
• Approaches are homology modeling and
protein threading
Physics-based approach
• Use only the laws of Physics to model folding
processes and resulting native structures.
• Aim to not use statistical energy functions or
secondary structure predictors.
Like Homology Modeling, Protein Threading.
• Now being combined with some database
information.
Physics-based approach – Advantages
• Can predict conformational changes
Induced-fit theory: important in computational drug discovery
• Predict conformational transitions
Maybe those based on environmental factors
• Design synthetic proteins
Foldable polymers for non-biological backbones
Induced-fit theory
Physics-based approach – Problems
• Inaccuracies in “force-fields”
• Really, really high computational requirements
At least right now
Empirical Force-Fields
Computational Cost
Physical time for simulation
10-4 Seconds
Typical time-step size
10-15 Seconds
Number of MD time steps
1011
Atoms in a typical protein and water simulation
32,000
Approximate number of interactions in force calculation
109
Machine instructions per force calculation
1000
Total number of machine instructions
1023
Planned supercomputer capacity in 2009: 1 petaflop
1015
One year to simulate folding of small protein
approximately…
Recent Advances
• 36-residue villin folded in 1μs
– Explicit solvent, initially unfolded
– Duan, Kollman (1998)
– 4.5A RMSD
• 20-residue Trp-cage folded in 92ns
– Implicit solvent
– Around 1A RMSD
• Folding@Home folded villin to 1.7A RMSD
Root Mean Square Deviation
If molecular orientation changes in arbitrary way, lRMSD or Least RMSD is
used to find optimal alignment using the Kabsch Algorithm or Quaternions
Problem Three
Unraveling Folding Kinematics
Some important ideas
• Afinsen’s Paradigm (1957)
All the information required to make a 3-D native structure is
contained in the amino-acid sequence
• Levinthal’s Paradox (1968)
If a protein sampled all possible conformations, time to reach
‘correct’ one would be more than age of universe
• Protein Sequence Space
With 20 amino acids as ‘alphabet’, how many theoretical
proteins are possible? What about evolution?
• Baldwin Conjecture
Understanding protein folding can lead us to devise better
algorithms to predict native structures from amino-acid sequences
Folding speed
• Two decades ago, could not measure anything
faster than few milliseconds.
• Technology exists now.
– Laser temperature-jump methods
– Mutational methods to identify amino-acids
controlling folding speed
– Förster resonance energy transfer (FRET) methods to
‘watch’ formation of contacts
– Hydrogen-exchange methods to ‘see’ structural events
– Extensive studies on protein models
Cytochrome c, barnase, chymotrypsin inhibitor 2
Now what about Levinthal’s speed
principle?
What we know
• Folding does not happen in a single
microscopic pathway
– “Funnel-shaped” landscapes
– Going from non-native state to native state is
different for each conformation of same sequence
• Folding processes are heterogenous
– Observations see averages and not distributions,
variations
What we don’t know
• How do folding rates change with specific
mutations?
• How to characterize kinetic heterogeneity?
– Single-molecule experiments
• Master-equation theories
What about Baldwin’s Conjecture?
Question
• How to design simulations which can arrive at
native state faster and more accurately than
Monte Carlo or molecular dynamics?
Need to know microscopic folding routes
A possible answer
• Zipping and Assembly (ZA)
– Proteins do not reach all their degrees of freedom
at the same time
– Fold over a range of timescales.
– Fast timescale (nano to pico): Small peptide pieces
explore conformations independently.
– Formed local structure “zips”, includes more
surrounding chain.
– Further assembly on slower timescale.
Does it work?
• ZA speeds up conformational searching.
• Physics-only models can find approximately
correct folds for 25-75 monomers.
– Ozkan et al, 2006
– Used AMBER96 force-field, implicit solvent
– Tested 9 proteins from PDB
– Eight were within 2.2A (avg.) RMSD
• Gives a good overall sense of folding
mechanism
Difficulty
Can we know the conformations the overall
protein does not search?
Important in understanding proteopathies and
hence drug design.
Conclusions
• Sophisticated problem
– Protein-protein interactions, cofactors, multidomain protein behavior, cytosolic interactions
and effects unknown .
• But some headway is being made
– Small proteins’ native structures and folding codes
are being determined accurately
– What we know is sufficient to design new proteins
and polymers (foldamers)
– Good contributions to novel drug discovery and
proteomics
• Good sense of Levinthal’s optimization puzzle
Questions
or
Comments?
Thank you
What happens if a polypeptide
does not fold properly?
Structure is related to function
• Resulting protein is rendered biologically,
functionally inactive
• Simpler forms can be degraded by cell machinery
• Amyloid accumulation (Proteopathies)
Alzheimer’s, Parkinson’s
• Can re-fold other normal proteins (Prions)
Creutzfeld-Jakob Disease
Sources
•
http://www3.interscience.wiley.com/journal/66000862/abstract?
CRETRY=1&SRETRY=0
•
http://opa.faseb.org/pdf/protfold.pdf
•
http://www.biozentrum.unibas.ch/~schwede/Teaching/BixIISS05/FR-HM.pdf
•
http://dasher.wustl.edu/bio5476/reading/curropstrbio-14-7604.pdf
•
http://arxiv.org/ftp/q-bio/papers/0402/0402039.pdf
•
http://www.biostat.jhsph.edu/~iruczins/presentations/ruczinsk
i.04.04.retreat.pdf