Day 9 part 2
๐งฌ Molecular Dynamics (MD) Simulation โ Theoretical Summary
โ๏ธ 1. Newtonโs Equations and Motion of Atoms
Molecular dynamics simulations are based on classical Newtonian mechanics.
- Each atom has:
- Position (x)
- Velocity (v) โ derivative of position
- Acceleration (a) โ derivative of velocity
- Motion is described by Force = mass ร acceleration
This means MD simulation becomes a huge system of ordinary differential equations because:
- For M atoms โ 3M position coordinates + 3M velocity coordinates
- So the system exists in a 6M-dimensional phase space
๐ Because this system is extremely complex, an analytical (exact) solution is impossible. Instead, MD uses numerical integration โ moving atoms step-by-step in time.
โฑ๏ธ 2. Time Step โ The Most Critical Parameter
MD simulations advance in small discrete time steps (ฮt).
Typical values:
- ๐ข ~2 femtoseconds โ standard protein simulations
- ๐ก ~4 femtoseconds โ very long or membrane simulations
Why is time step important?
If the time step is:
โ Too small
- Simulation becomes very slow
- You may never reach biologically relevant timescales (ยตs or ms)
โ Too large
- You may miss important events like collisions
- System can become numerically unstable
- Atoms may โjumpโ outside the simulation box โ simulation โexplodesโ
Thus, MD requires a trade-off between accuracy and computational speed.
๐ธ 3. Leapfrog Integration Algorithm
To propagate motion efficiently, MD often uses the leapfrog algorithm.
Concept:
- Velocities and positions are calculated at offset time points
- Values โjump overโ each other like a frog hopping
Advantages:
- Time-symmetric
- Numerically stable
- Accurate for long simulations
This makes it one of the most widely used integration methods in MD.
๐ก๏ธ 4. Atoms Are Always Moving (Even at Low Temperature)
Important physical concept:
- Atoms are never static
- Even near 0 K โ quantum and thermal motion still exist
Thus:
- Proteins do not sit permanently in minimum energy
- They constantly fluctuate around it
Over long simulation time, atomic configurations follow the:
๐ Boltzmann Distribution
This distribution describes:
The probability of a conformation as a function of its potential energy.
- Low-energy conformations โ frequent
- High-energy conformations โ rare
If MD samples all conformations, we can calculate thermodynamic properties identical to experiments.
But reaching full sampling requires very long simulations.
๐ฐ 5. MD vs Monte Carlo Sampling
Monte Carlo (MC)
- Samples conformations randomly
- Immediately shows distribution
- Good for thermodynamic properties
Molecular Dynamics (MD)
- Follows realistic time evolution
- Captures:
- Fast motions
- Dynamic pathways
- Mechanistic transitions
Thus:
- MC โ better statistics
- MD โ better physical realism and motion information
๐ 6. Energy Conservation and Thermostats
Total system energy:
E_ = E_ + E_
However:
- Real atomic collisions are not perfectly elastic
- Energy is gradually lost
Therefore MD uses:
๐ Heat bath / thermostat
Functions:
- Adds energy if system cools
- Removes energy if system overheats
- Maintains constant temperature
Energy curves therefore fluctuate around a mean value, not perfectly flat.
๐ 7. Equilibration Phase
At simulation start:
- All atoms may have identical velocities
- This is physically unrealistic
During equilibration:
- Velocities redistribute according to mass
- System relaxes to correct temperature distribution
- Artifacts from starting conditions disappear
Only after equilibration โ production simulation begins.
๐ง 8. Solvent Representation
Explicit Solvent
- Real water molecules simulated
- Most accurate
- Very computationally expensive
- Can introduce artifacts (water clustering, ion shells)
Implicit Solvent
- No water molecules
- Environment treated mathematically like water
Advantages:
- Faster
- Fewer atoms
Disadvantages:
- Loss of specific hydrogen-bond interactions
- Reduced realism near protein surface
๐ฆ 9. Periodic Boundary Conditions (PBC)
To avoid atoms leaving the system:
- Simulation box is surrounded by copies of itself
When atom exits one side:
- It re-enters from opposite side
Benefits:
- Constant particle number
- No artificial wall effects
- Mimics infinite bulk environment
โก 10. Computational Limits of MD
Challenges:
- Very small time steps required
- Structural changes may occur on ms scale
- Huge number of steps needed
Solutions:
- GPU acceleration
- Parallel computing
- Algorithm improvements
๐ 11. Force Fields โ Semi-Empirical Models
Force fields are:
- Parameter tables describing atomic interactions
- Based on:
- Physics equations
- Experimental data
Thus MD is semi-empirical.
โ ๏ธ Must use scientific judgement:
- Simulations may produce physically unrealistic minima
- Example: alkane chain passing through benzene ring (artifact)
Always check: ๐ Does the simulation result make chemical sense?
โ 12. What MD Cannot Do
Classical MD limitations:
- Cannot form or break covalent bonds
- Cannot change protonation states dynamically
- Cannot simulate enzyme reactions directly
Reason:
- Electronic structure is fixed at simulation start
- Born-Oppenheimer approximation assumed
To model reactions โ need QM/MM methods (quantum mechanics + molecular mechanics).
๐ฅ๏ธ 13. Software and Acceleration
Common MD packages:
- GROMACS
- AMBER
- CHARMM
Visualization:
- VMD (Visual Molecular Dynamics)
Speed improvements:
- Parallelization
- GPU (CUDA cores especially important)
โญ Key Conceptual Takeaway
Molecular dynamics simulations:
- Provide time-resolved molecular motion
- Sample conformational landscapes
- Require careful parameter choices
- Balance accuracy vs computational feasibility
- Must always be interpreted with chemical and physical intuition