Conformational Ensembles from Experimental Data and Computer Simulations

Conformational Ensembles from Experimental Data

and Computer Simulations

Sunday Speaker Abstracts

Modelling Conformational Ensembles from Small Angle Scattering

Jill Trewhella

, Wojciech Potrzebowski

, Ingemar André

Biochemistry & Structural Biology, Lund University, Lund, Sweden,

The University of

Sydney, Sydney, Australia.

We have applied a Baysian approach to ensemble modelling against solution Small-Angle

Scattering (SAS) and NMR chemical shift data for our two domain NMR solution structure of

ΔmC2 (PDB:2KDU Michie et al. 2016, Structure 24, 2000) from cardiac Myosin Binding

Protein C (cMyBP-C). ΔmC2 has two folded domains linked by 7 highly flexible amino acids

that are the surprisingly also highly conserved and include severe disease-linked mutation sites.

We postulate it to be a polymorphic binding domain that interacts with multiple proteins to

regulate muscle action in the sarcomere.

The small-angle scattering (SAS) from proteins in solution samples the ensemble average of the

randomly oriented structures, and ensemble modelling for proteins with flexible regions against

SAS data is increasingly popular. However, the smooth SAS profile can typically be defined by

as few as 10-15 points, and the ensemble model has many more degrees of freedom. Typically, a

very large ensemble (10,000 or more) is generated within some constrained set, and a population

weighted sub-set of structures is identified that predicts a profile that best-fits the data.

Representative structures are selected based on clustering analysis to aid in visualizing the nature

of the ensemble, but their accuracy and what minimal set is justified by the data are outstanding

questions.

The alternate Bayesian approach assigns a posterior probability for the population weight of each

structure in an ensemble. As a result, uncertainty in the parameters of the ensemble can be

quantified so that inferences can be made using standard statistical methods. We will present the

results of our Bayesian approach using SAS or NMR chemical shift data alone, and SAS plus

chemical shift data, and the effects of the quality and size of structural library on the selected

models and their populations.