Biophysical Flashcards
What are the main types of protein (5)
Enzymes, structural, regulatory, signalling, defensive
What are the 2 techniques used to determine protein structure
NMR and x-ray crystallography
How does x-ray crystallography work
Phases determined, Fourier Transform gives diffraction pattern into electron density map. Molecular model built into density and refined. The structure is an average over molecules and time of experiment.
What parts of proteins are sometimes not seen in structure determination and why?
Termini and loops: dynamic and static disorder
What was the first crystal structure of and what did it show us?
Myoglobin; complexity, lack of symmetry and regularity
What is a con of crystallography of proteins
They don’t show bend/flex therefore can’t study in action and don’t fully represent real structure or function because the transition state is the important part of the reaction
Challenges in biology (3)
Protein folding: prediction of quaternary structure from primary
Enzymes and design of biomimetic catalysts
Ligand binding: can we predict binding affinity for drugs
What does dynamics drive (3)
Association (eg membrane insertion)
Folding and conformation
Chemical reactions (eg ligand binding)
Pros of NMR vs crystal structures (3)
Only procedure for disordered and denatured states at atomic resolution in solution
Natural protein dynamics
15N, 13C, 1H –> protein
Con of NMR
High conc required
How do we obtain an ensemble of structures from different NMR techniques
COSY gives peaks for a spin system. NOESY gives peaks for proximity where intensity decreases with increased separation. Get distance and angle restraints which builds into an ensemble of structures that satisfy restraints.
Pro of crystallography
Higher resolution
How does ultrafast x-ray imaging work
Causes Coulombic explosion (atom not useful, blown up after few pulses), measures energy.
Showed breathing motion around Heme
What are the basic principles of molecular dynamics
Atomic motion is governed by the forces atoms feel from environment, which arise from electrostatic interactions i.e. potential energy
What are the 2 cons of Molecular Dynamics simulations
Requires significant computational power and scales poorly
What is Molecular Mechanics
evaluation of V(q) in big molecules, used in MD simulations for investigation of structure and folding, DNA, biomolecules
Why is MM time consuming
It uses maaaany parameters and you have to fit them all. For example propane: 18 torsions, 27 non bonded interactions, 73 energy terms
What are the 2 important MM programs
AMBER, CHARMM
What is a potential energy landscape
A 3(N-6)D surface plot giving toplogical features arising from V(q). Many metastable local minima and very few/one global minimum
Why do all parts of a molecule move all the time
zero point energy, thermal energy
What approximations do protein MD rely on (2)
- Born-Oppenheimer: only explore a single electronic state
- Nuclei are classical objects: can be treated with Newton’s equations
Applications of MD to biology (4)
- Identifying and analysing functionally important structural changes in proteins
- Studying how ion channels work
- Drug design
- Protein folding
Describe MD process
1) specify position q, velocities v, forcefield V(q), timestep delta t
2) calculate forces F = -(dV(q)/dq) and acceleration a = F/m
3) Update position
4) update velocities
5) move time forward, feed into 2
Describe procedure for setting up MD simulation of protein
- Choose starting (crystal) structure
- Build in missing parts eg disordered loops/termini
- Assign protonation states to titratable residues
- Add H atoms (not observed in crystal structures)
- Add solvent (phase protein in water box)
- Minimise E of system (e.g. by steepest descents minimisation: relieve strain and remove bad contacts)
Procedure for running MD simulation of a protein
Run until system reaches equilibrium (average important values are stable), start to calculate properties
What is a thermostat
Specific temperature distribution maintains constant ensemble. it is an algorithm to enforce
What is the most expensive part of an MD simulation
The evaluation of the dV(q)/dq vector is the most expensive part of MD
What are the 2 ways to evaluate dV(q)/dq
1) Solve TDSE HY(r,q) = V(q)Y(r,q)
- v accurate and expensive
2) Use MM forcefields: computationally efficient, less accurate because cannot capture chemical processes which involve breaking/making bonds
How does Quantum Mechanics/Molecular Mechanics (QM/MM) work?
Small QM region eg active site of an enzyme with surrounding MM for protein and solvent. The two regions interact by Van der Waals and electrostatics. QM atoms ‘feel’ MM charges and are polarised by them.
3 different forms of QM
ab initio, DFT, semi-empirical
3 things to consider when picking a method
- Can it answer the question
- Balance of accuracy and expense
- Understand strengths and weaknesses of results, judge critically
What is a general description of MD
It is a solution to Newton’s equations of motion in the form of a step-by-step algorithm of forces positions and velocities for a small timestep.
How does protein folding occur in terms of the potential energy landscape
Funnel shape guides dynamics of protein and helps find shape for function; folding occurs via an ensemble of potential microscopic trajectories. The low energy structure is kinetically trapped in shape for job
What is the protein folding problem? (3)
1) physical code by which an amino acid sequence becomes a native structure?
2) How can proteins fold so fast?
3) Can we devise an algorithm to predict structure from sequence?
What is Levinthal’s Paradox?
2 torsional angles 0 and Y; 99 peptide bonds –> 198 0 and Y angles, each can take 3 conformations therefore 3^198 conformations and that is just the backbone!
Not strictly true; just because one value of Y is something doesn’t mean another angle can’t be Y. But a decent approximation.
What do simulations of protein folding give us?
- timescale of protein folding
- pathways with defined order
- allow study of unfolded state (difficult experimentally)
Seen in folding simulations, what parts form quickly/slowly?
Really quick formation of secondary structure, native contacts form more slowly. Helices and sheets form quickly but are out of place, later fall into place
How does use of Graphical Processing Units (GPUs) help simulation
They are bad at decision making eg emails but good at maths, fast enough for interactive protein simulation; significantly accelerate simulation.
Give an example of GPUs have helped research
Drug resistant mutants of influenza neuraminidaze (catalyses repair of bactera cell wall) - previously 1 us required 3 months with a CPU. 8 ns/day with one GPU, 18 us a month of dynamics; enough to see multiple binding and unbinding events.
How does folding@home work?
Distributed computing from master splits task. Relies on short dynamics simulation. Master splits task into bits, each worker simulates part and gets a rate coefficient (results) and feeds back to master.
Rate coefficients used to build transition matrix M –> solve n(t) by matrix or stochastic approach
What is a Markov Model?
Way of modelling how a system evolves in time based on kinetic parameters ie rate coefficients
What does a Markov model assume?
Future states of a system depend only on current state and not on previous events eg as in rate coefficients
What are the two approaches to solving for n(t) in a Markov Model
matrix and stochastic approach
How does the matrix approach to solving for n(t) work in a Markov model and why can it sometimes be inaccurate?
It is an exact and desired procedure. Solution of a specific form of eigenvectors.
Where matrix v large over range of timescales, matrix diagonalisation is long and results can be wrong due to numerical errors (rounding)
How does the stochastic approach to solving for n(t) work in a Markov model
Consider possibilities for timestep delta t, consider magnitude of rate coefficients which are proportional to relative probability of event
What is the Kinetic Monte Carlo Procedure
Give 4 cons and 1 pro
Move to R2 –> stay in R2 –> move to P –> converge on limit
Takes long for topologies with lots of metastable states or deep wells
Statistical
Requires several simulations and results average until convergence
Not subject to numerical error
What is a homology model
a structure for a known protein sequence built based on its similarity to the experimentally known structure of a related homologous protein, based on aligning sequences
What mathematical tool fully specifies time-dependent dynamics?
The Markov Transition Matrix; in principle, we have a full description of dynamics if we can construct all the transition probabilities in M
How are transition matrix elements in a Markov model usually denoted?
THey are unimolecular rate coefficients, /s
What do PEL barriers correspond to in protein folding?q
Conformational changes; small but many
What do PEL barriers correspond to in enzymology?
Bond breaking/making, much larger
Why is enzyme simulation difficult?
It is not possible to sample enzyme binding with MD the very large barriers
Why are enzymes so efficient and specific?
They lower Ea by about 10^10 times acceleration, can reach 10^16; very specific and efficient
How long would decarboxylation take without enzymes at 25 C
2.3 billion years
What is the enzyme complimentary to
the transition state, or “activated molecule”
How does the enzyme lower the barrier to reaction
It binds the transition state more strongly than the substrate
Implications of transition state stabilisation by enzymes (3)
1) transition state analogues (stable molecules which represent transition state) should be excellent enzyme inhibitors and maybe drugs
2) Designed transition state analogues are potent inhibitors of purine nucleoside phosphorylase
3) Use electrostatic PE surfaces: design one complementary to transition state; complementarity comes from electrostatic interactions
What is the approximate efficiency of an abzyme
kcat/kuncat ~ 10^6 … our understanding of how enzymes work is incomplete
What is the recent proposal of why we don’t understand how enzymes work?
Their function relies as much on dynamics as structure
- Heavy rearrangement rxns neglected, H transfer important and large as H is light and rxns can tunnel. H bonding has a non-negligible de Broglie wavelength
What is the Kinetic Isotope Effect for Aromatic Amine Dehydrogenase?
50 - 60 kH/kD
What are the two strategies to model tunnelling in MD
1) Rare-event sampling: map free energy surface along reaction coordinate G(lamda), calculate kappa(T) based on G(lamda) curvature
- classical approach
2) Free-energy mapping: quantum effects give kappa(T)
- semi-classical approach
- more expensive
What does H transfer in enzymes show
KIEs difficult to explain using 1-state transition state theory
