Molecular dynamics (MD) simulations compute the time evolution of a biomolecular system by numerically integrating Newton's equations of motion for every atom, using empirical force fields (AMBER, CHARMM, OPLS) that describe bonded and non-bonded interactions. Starting from an experimental structure, MD reveals protein dynamics — conformational fluctuations, domain motions, ligand binding/unbinding, and allosteric transitions — at atomic resolution and femtosecond time resolution. Modern simulations routinely reach microsecond to millisecond timescales (with specialized hardware like Anton reaching beyond), capturing functionally relevant conformational changes inaccessible to experimental methods. MD also enables free energy calculations (binding affinities, mutational effects) that connect structure to thermodynamics.
Experimental structural biology provides snapshots — a crystal structure is one conformation, a cryo-EM map captures a few discrete states. But proteins are dynamic molecular machines whose function depends on motion: enzymes flex to accommodate substrates, receptors change shape to transmit signals, and channels open and close gates. Molecular dynamics simulation bridges the gap between static structures and dynamic function by computing how every atom in the system moves over time.
The physics is classical mechanics. Each atom is treated as a point mass interacting with other atoms through a force field — a set of mathematical functions and parameters that describe bonded interactions (bond stretching, angle bending, torsion rotation) and non-bonded interactions (van der Waals attraction/repulsion, electrostatic attraction/repulsion between partial charges). The force on each atom is computed from the force field, Newton's second law (F = ma) gives the acceleration, and numerical integration (typically the Verlet algorithm with a 2-femtosecond time step) advances the positions and velocities. Repeating this for billions of time steps generates a trajectory — a movie of the molecular system's evolution at atomic resolution.
Modern MD routinely simulates systems of 100,000-1,000,000+ atoms (the protein, surrounding water molecules, ions, and sometimes a lipid membrane) for microsecond timescales. The force field accuracy has been refined over decades, and current-generation force fields (AMBER ff19SB, CHARMM36m) reproduce experimental observables (NMR relaxation, J-couplings, folding thermodynamics) with impressive accuracy for many systems. The timescale frontier has been pushed by specialized hardware: D.E. Shaw Research's Anton computer has achieved millisecond simulations, directly observing protein folding, drug binding kinetics, and allosteric transitions that were previously inaccessible.
The applications of MD span structural biology. Conformational dynamics: simulations reveal the full range of motions a protein undergoes, identifying hinge regions, breathing motions, and transient states that are invisible to static structural methods. Drug discovery: free energy perturbation (FEP) calculations predict how chemical modifications to a drug candidate affect binding affinity, guiding medicinal chemistry optimization. Mechanism: simulations of enzyme active sites reveal the catalytic mechanism at atomic detail, including the role of dynamics in positioning catalytic residues. Membrane proteins: MD simulates proteins in lipid bilayers, revealing how the membrane environment affects channel gating, transporter mechanism, and receptor activation. The combination of MD simulation with experimental structural data creates a comprehensive picture of biomolecular structure and dynamics that neither approach could achieve alone.