SASSIE: Modelling AUC & SAS Data Using Atomistic Simulations

25th July 2017

David Wright & Emre Brookes

Molecular modelling and simulation: What is it?

• Theoretical and computational methods used to model or mimic the behaviour or properties of molecules
• Atomistic representation of molecules
• Encode chemistry
• Use classical or quantum mechanics to describe interactions

Molecular modelling and simulation: What is it good for?

• Understanding experimental results
• Combining information from multiple experiments
• Providing atomistic explanations for higher level observations

+

+

What do we need to model AUC & SAS experiments?

• Initial models capturing known chemistry
• Variety of physically plausible conformations
  • Global structure
  • Domain level rearrangements
• Atomistic models we can analyse

Molecular Dynamics

Classical (Newtonian) dynamics

F = m a

Forcefield description of interactions

F=−∇ U

Molecular Dynamics

MD

Experiment

Model system with N particles, solve F = ma until properties do not change (equilibrate) then you “measure” (i.e. average a property) until data converge

Sample in instrument, measure over time ... measure longer until data converge.

Molecular Dynamics

MD and experiment can both suffer from the similar issues

Molecular Dynamics

Requirements

• Initial structure
• Forcefield

Output

• Trajectory of coordinates

Molecular Dynamics

• Time step determined by fastest motion
• Usually X-H bond
• Even fastest supercomputers allow only microsecond simulations in general

Monte Carlo Simulation

• Vary system
• Evaluate energy
• Keep new structure if
  1. ​More energetically favourable or
  2. With probability related to energy change
• Repeat first step

Dihedral Angle Monte Carlo Simulation

Frenkel and Smit, Understanding Molecular Simulation

Dihedral Angle Monte Carlo Simulation

• MC with all degrees of freedom varied at least as slow as MD
• "Freezing" some degrees of freedom potentially gives a huge speedup
•  Sample only the dihedral potential using the Metropolis criterion
  • Dihedral Angle Monte Carlo

Dihedral Angle Monte Carlo

8 cores

13 days

1 core

15 mins

• Rapid generation of ensemble
• Sampling Ramachandran obviously limiting
• Use as hypothesis to test against data

SASSIE

Simple unified interface to tools facilitating:

• Initial model building/preparation
• Simulation
  • ​Dihedral angle Monte Carlo
  • Molecular dynamics
  • Torsion angle MD
• Calculation
  • ​SAS curves
  • AUC profiles
• ​Comparison to experimental data

https://sassie-web.chem.utk.edu/sassie2/

Further Reading

Full three day training course (today is similar to day 2):

https://sassie-web.chem.utk.edu/training/uk_2017/main.html

MD

A. Leach, Molecular Modelling: Principles and Applications

J. D. Durrant & J. A. McCammon, Molecular dynamics simulations and drug discovery, BMC Biology, 2011, 9:71, DOI: 10.1186/1741-7007-9-71

Dihedral Angle Monte Carlo

J. E. Curtis et al, SASSIE: A program to study intrinsically disordered biological molecules and macromolecular ensembles using experimental scattering restraints, Computer Physics Communications, 2012, 183:2, DOI: 10.1016/j.cpc.2011.09.010

Torsion Angle MD

W. Zhang et al, Combined Monte Carlo/torsion-angle molecular dynamics for ensemble modeling of proteins, nucleic acids and carbohydrates, J Mol Graph Model, 2017, 73, DOE: 10.1016/j.jmgm.2017.02.010