Lecture 13: Conformational Sampling: MC and MD Dr. Ronald M. Levy Contributions from Mike Andrec and Daniel Weinstock Statistical Thermodynamics.

Slides:



Advertisements
Similar presentations
Time averages and ensemble averages
Advertisements

PRAGMA – 9 V.S.S.Sastry School of Physics University of Hyderabad 22 nd October, 2005.
Simulazione di Biomolecole: metodi e applicazioni giorgio colombo
Statistical mechanics
Molecular dynamics in different ensembles
Biological fluid mechanics at the micro‐ and nanoscale Lecture 7: Atomistic Modelling Classical Molecular Dynamics Simulations of Driven Systems Anne Tanguy.
Advanced Molecular Dynamics Velocity scaling Andersen Thermostat Hamiltonian & Lagrangian Appendix A Nose-Hoover thermostat.
Molecular Biophysics III – dynamics
Molecular Dynamics at Constant Temperature and Pressure Section 6.7 in M.M.
Transfer FAS UAS SAINT-PETERSBURG STATE UNIVERSITY COMPUTATIONAL PHYSICS Introduction Physical basis Molecular dynamics Temperature and thermostat Numerical.
Survey of Molecular Dynamics Simulations: Week 2 By Will Welch For Jan Kubelka CHEM 4560/5560 Fall, 2014 University of Wyoming.
© copyright 2013-William A. Goddard III, all rights reservedCh120a-Goddard-L06 Ch121a Atomic Level Simulations of Materials and Molecules William A. Goddard.
Monte Carlo Methods and Statistical Physics
Molecular Dynamics. Basic Idea Solve Newton’s equations of motion Choose a force field (specified by a potential V) appropriate for the given system under.
Lecture 14: Advanced Conformational Sampling
The Calculation of Enthalpy and Entropy Differences??? (Housekeeping Details for the Calculation of Free Energy Differences) first edition: p
MSEG 803 Equilibria in Material Systems 8: Statistical Ensembles Prof. Juejun (JJ) Hu
Graphical Models for Protein Kinetics Nina Singhal CS374 Presentation Nov. 1, 2005.
Molecular Dynamics Classical trajectories and exact solutions
Joo Chul Yoon with Prof. Scott T. Dunham Electrical Engineering University of Washington Molecular Dynamics Simulations.
Bioinf. Data Analysis & Tools Molecular Simulations & Sampling Techniques117 Jan 2006 Bioinformatics Data Analysis & Tools Molecular simulations & sampling.
Advanced methods of molecular dynamics Monte Carlo methods
1 Statistical Mechanics and Multi- Scale Simulation Methods ChBE Prof. C. Heath Turner Lecture 11 Some materials adapted from Prof. Keith E. Gubbins:
Introduction to (Statistical) Thermodynamics
Molecular Dynamics Simulations
Room 2032 China Canada Winnipeg Manitoba.
Free energies and phase transitions. Condition for phase coexistence in a one-component system:
Javier Junquera Molecular dynamics in the microcanonical (NVE) ensemble: the Verlet algorithm.
Molecular Dynamics Simulation Solid-Liquid Phase Diagram of Argon ZCE 111 Computational Physics Semester Project by Gan Sik Hong (105513) Hwang Hsien Shiung.
1 CE 530 Molecular Simulation Lecture 17 Beyond Atoms: Simulating Molecules David A. Kofke Department of Chemical Engineering SUNY Buffalo
Molecular Dynamics A brief overview. 2 Notes - Websites "A Molecular Dynamics Primer", F. Ercolessi
The Nuts and Bolts of First-Principles Simulation Durham, 6th-13th December 2001 Lecture 18: First Look at Molecular Dynamics CASTEP Developers’ Group.
Minimization v.s. Dyanmics A dynamics calculation alters the atomic positions in a step-wise fashion, analogous to energy minimization. However, the steps.
Basic Monte Carlo (chapter 3) Algorithm Detailed Balance Other points.
Basics of molecular dynamics. Equations of motion for MD simulations The classical MD simulations boil down to numerically integrating Newton’s equations.
Force Fields and Numerical Solutions Christian Hedegaard Jensen - within Molecular Dynamics.
1 CE 530 Molecular Simulation Lecture 6 David A. Kofke Department of Chemical Engineering SUNY Buffalo
P T A typical experiment in a real (not virtual) space 1.Some material is put in a container at fixed T & P. 2.The material is in a thermal fluctuation,
Monte Carlo Methods in Statistical Mechanics Aziz Abdellahi CEDER group Materials Basics Lecture : 08/18/
Statistical Mechanics and Multi-Scale Simulation Methods ChBE
Understanding Molecular Simulations Introduction
Advanced Molecular Dynamics Velocity scaling Andersen Thermostat Hamiltonian & Lagrangian Appendix A Nose-Hoover thermostat.
7. Lecture SS 2005Optimization, Energy Landscapes, Protein Folding1 V7: Diffusional association of proteins and Brownian dynamics simulations Brownian.
1 CE 530 Molecular Simulation Lecture 23 Symmetric MD Integrators David A. Kofke Department of Chemical Engineering SUNY Buffalo
Molecular Modelling - Lecture 2 Techniques for Conformational Sampling Uses CHARMM force field Written in C++
Incremental Integration of Computational Physics into Traditional Undergraduate Courses Kelly R. Roos, Department of Physics, Bradley University Peoria,
ChE 452 Lecture 25 Non-linear Collisions 1. Background: Collision Theory Key equation Method Use molecular dynamics to simulate the collisions Integrate.
Molecular dynamics (1) Principles and algorithms.
7. Metropolis Algorithm. Markov Chain and Monte Carlo Markov chain theory describes a particularly simple type of stochastic processes. Given a transition.
Monte Carlo method: Basic ideas. deterministic vs. stochastic In deterministic models, the output of the model is fully determined by the parameter values.
An Introduction to Monte Carlo Methods in Statistical Physics Kristen A. Fichthorn The Pennsylvania State University University Park, PA
Molecular dynamics (2) Langevin dynamics NVT and NPT ensembles
Review Session BS123A/MB223 UC-Irvine Ray Luo, MBB, BS.
1 Statistical Mechanics and Multi- Scale Simulation Methods ChBE Prof. C. Heath Turner Lecture 17 Some materials adapted from Prof. Keith E. Gubbins:
CS Statistical Machine learning Lecture 25 Yuan (Alan) Qi Purdue CS Nov
Statistical Mechanics and Multi-Scale Simulation Methods ChBE
Basic Monte Carlo (chapter 3) Algorithm Detailed Balance Other points non-Boltzmann sampling.
Lecture 14: Advanced Conformational Sampling Dr. Ronald M. Levy Statistical Thermodynamics.
Computational Physics (Lecture 11) PHY4061. Variation quantum Monte Carlo the approximate solution of the Hamiltonian Time Independent many-body Schrodinger’s.
Diffusion over potential barriers with colored noise
Dynamical correlations & transport coefficients
Overview of Molecular Dynamics Simulation Theory
Fundamentals of Molecular Dynamics Simulations
Molecular Modelling - Lecture 3
Molecular Dynamics.
Dynamical correlations & transport coefficients
Teaching Assistant: See website
Advanced Molecular Dynamics
Common Types of Simulations
Presentation transcript:

Lecture 13: Conformational Sampling: MC and MD Dr. Ronald M. Levy Contributions from Mike Andrec and Daniel Weinstock Statistical Thermodynamics

Importance Sampling and Monte Carlo Methods Energy functions are useless without sampling methods Knowing the energy of every point in a high-dimensional phase space is essential but not terribly informative Thermodynamic quantities are averages over an ensemble over the entire phase space We are often interested in the distribution of certain quantities (e.g. radius of gyration) averaged over all of the “uninteresting” degrees of freedom (aka “potentials of mean force”)

Thermodynamic quantities are averages over an ensemble over the entire phase space This could be evaluated by numerical integration over a grid, or by generating random points uniformly over phase space and estimating the integral by MC integration: In fact, the complete phase space may not be needed, since the velocity contributions can often be accounted for analytically. Then, we only need to consider the potential energy of conformational degrees of freedom.

These methods are generally hopeless for molecular systems: N is huge, Q is unknown, and most points in a uniform sampling have a very small value of the integrand. Frenkel & Smit (2002) Understanding Molecular Simulations, 2nd Ed., Academic Press

Importance Sampling by Markov Chain Monte Carlo Given a current point  i in configuration space, choose a subsequent point  i+1 with transition probability  (  i,  i+1 ) that depends only on  i. To get the correct sampling, it is sufficient that the transition probabilities satisfy microscopic reversibility:  (  i )  (  i,  i+1 ) =  (  i+1 )  (  i+1,  i ), or Want to produce conformations  distributed according to i j (flux i to j = flux j to i = equilibrium)

i j Proposal probability (probability of picking move) Acceptance probability (probability of accepting move once it has been selected) one valid choice: If 4 (symmetric MC scheme)

The “classic” Metropolis algorithm Pick a degree of freedom x Displace x by a uniformly distributed random number in range ±  Calculate the potential energy difference between the current state i and the proposed displaced state j Accept the move if Otherwise draw random number and accept if “Reject” does not mean “omit”… i j

Connections between microscopic information such as atomic positions and velocities and macroscopic observables is through statistical mechanics. In statistical mechanics, averages are defined as ensemble averages However, in MD simulations, we calculate time averages Molecular Dynamics (MD) Ergodic Hypothesis

1956: Alder and Wainwright MD method developed to study interactions of hard spheres 1964: Rahman first simulation with realistic potential - liquid Argon 1971: Stillinger and Rahman first simulation of realistic system - liquid water 1977: McCammon, Gelin, and Karplus first protein simulation - BPTI Historical Background

System of N particles Positions of the N particles, Velocities of the N particles Energy(E) of the system Kinetic Energy Potential Energy, V( r) System Temperature Microcanonical ensemble (constant NVE) closed system - no energy enters or leaves energy conservation used to check MD algorithm

The forces are complicated functions of the coordinates, non-linear functions of position, so the set of 3N coupled differential equations cannot be solved analytically. Newton’s Equations of Motion

The integrator is the heart of an MD algorithm Given molecular position, velocities and other dynamic information at time t, we attempt to obtain the positions, velocities, etc. at a later time t+δt to a sufficient degree of accuracy Numerical Integration of the Equations of Motion Finite difference method: Not very accurate, leads to divergences unless δt is made very small.

Allen and Tildesley. Computer Simulations of Liquids Better O( δt 2 ) algorithms: velocity Verlet, position Verlet, leap frog, predictor-corrector,... ?

The Liouville operator For any property A : q-component propagates coordinates in time: p-component propagates momenta in time:

Trotter expansion In generalBecause L q and L p don't commute However if δt is sufficiently small: Trotter

rRESPA Defines integrator made of P steps each made of 3 operations: 1. Propagate momenta to δt/2 : 1. Propagate momenta to δt :

rRESPA equivalent to velocity Verlet Most other integrators can be derived using different forms of short time expansions of the Liouville propagator

Choice of Time Step Time Step should be small enough for trajectories to be close to exact Energy Conservation is used as a criteria for choosing the time step signals good energy conservation We want to use a time step that minimizes computational time, while maintaining Energy Conservation The faster time scales control the time step δt to use. To integrate vibrations need to choose δt much smaller than 1fs

Multiple time step rRESPA Decompose total force in a fast component (covalent interactions - inexpensive) and a slow component (non-bonded interactions - expensive): Short time propagator: Then break up inner fast propagator using a shorter time step: Fast forces are applied N times in the inner loop allowing larger time step in outer loop.

Canonical ensemble (constant N,V,T) There are different methods for constant temperature MD Andersen velocity resampling Nosé-Hoover thermostat (extended system) the system is coupled to extra degrees of freedom which simulate heat bath. Original system is canonical, extended system is microcanonical Langevin thermostat. Periodically atomic velocities are stochastically perturbed based on friction coefficient and relaxation time. Berendsen thermostat (velocity rescaling – non canonical) Constant Temperature Molecular Dynamics

Typical Organization of a MD simulation 1)Energy minimization 2)Thermalization 3)Equilibration 4)Production 5)Trajectory Analysis