Biomolecular Modelling and Simulation

Julia M Goodfellow, Birkbeck College, University of London

The Chemical Interface:Simulation and Biodesign

Example: Application of commonly used algorithm is that of molecular dynamics i.e. iterative solution of differential equations describing atomic motion.

•One bottleneck is cpu

•Second bottleneck is coordinated storage and analysis.

WrongWrong

RightRight WrongWrong

DisastrousDisastrous

AMYLOIDAMYLOID

PROTEIN MISFOLDING LEADS TO DISEASE

Human Lysozyme 2 microglobulin Transthyretin

eye-lens crystallins p53 - cancer

Human Lysozyme Two domains: and. 130 residue

enzyme Mutation causes repulsion between

distal loop and hairpin of domain. Amyloid precursor has been suggested

to be formed from destabilised domain. H-bond networks change.

D67H and I56T lead to non-neuropathic amyloidosis

Results: Unfolding rates

• Results agree with plots from native contacts plots whereby the number of native residue contacts decreases slightly faster in the mutant in all trajectories.

• Indication that mutant moves faster away from native structure

• RMSD shows that mutant unfolds slightly faster than wild type.

• Indication that more susceptible

• Clustering is used to identify favourable partially folded conformations

• Most conformations have lost most of their secondary structure.

• Highly populated conformational states of the mutant and a few low populated of wild type have distorted β domain

• The distortion is more evident in the area of the domain interface (Ile 56).

Results: Clustering

GENERALISE PROBLEM

1 ns simulation on 40,000 atom protein 20 days on typical single processor

Aiming for simulations of 10 trajectories of 10 ns = 100 ns

UK community say 10 groups of 10 people each studying 5 related systems

total days cpu = 20 x 100 x 100 x 5 = 1 x 106 days cpu

= 3000 PCs for a year.

1.5 Gb storage per 1 ns simulation = 70 terabytes storage per year.

PB electrostatics with conformational change

Lowering of pH results in break up of tetramer and changes to the monomer structure for transthyretin

Would like to be able to combine modelling of changes in conformation with changes in pH

45,000 atoms with solvent

Tyr 116Tyr116

His88

His90Glu92

Prototype GRID

cpu* disk archive

Centre with strong link to particle physics community

64 cpu 64 cpu 64 cpu 64 cpu

Other centresin future

GRID2 for Data Analysis

Distributed data - often on tape i.e. not on line

Oxford

BirkbeckBirmingham

York Southampton

Nottingham

Metadata

REST of WORLD

First Stage

Metadata -

• Define what has been done by whom

• is raw data available on line?

• make this information available to all

• Using Grid tools to allow sub-group (initially) to access raw data

Second stage

•Do we trust each others data?

•Developing a ‘kite’ mark for quality or resolution of data

• What are we going to use to do this ?

Third Stage

Developing analysis tool box ( not reinventing the wheel).

How are these to be used?

Testing of Grid tools - where do we run the analysis? Do we move ‘code to data’ or ‘data to code’? This may involve sharing of our servers for running programmes.

Where do we want to be?

• Sharing data with other modellers

• Integrating simulation data with other data

• Presenting data so it can be used/accessed by non-modellers

• Having all singing/all dancing database

• Using any spare capacity within our group for number crunching

We shall never get people whose time is money to take much interest in atoms.

Samuel Butler 1835-1902

Funding:

BBSRC, EPSRC

Wellcome Trust

AICR

Acknowledgements

George Moraitakis Mark Sansom - Oxford

Delphine Flatters Oliver Smart - Birmingham

Spiros Skoulakis Jonathan Essex - Southampton

Isofina Pournara Leo Caves - York

Andy Purkiss Charlie Naughton - Nottingham

Thomas Matthews Paul Jeffrey (Oxford)

David Boyd & Paul Durham (CLRC)

Electrostatic stability of wild type and mutant transthyretin oligomers

90°

90°

pH induced changes

V30M

T119M

CPU timings

~17,000 atoms using Gromacs 2.0 -

8 processor Beowulf Cluster - 6 ns 12.5 days

27,000 atoms using AMBER 6

4 processor Beowulf Cluster - 3 ns 23 days

pK1/2 values monomer

His31 5.1

His56 4.8

His88 3.9

His90 3.6

Glu54 2.1

Tyr116 9.3

Glu92 0.3

His88 -5.3

His90 -7.6

Tyr116 -1.7

Glu92 -5.3

Tyr116

Glu92

pK1/2 values dimer

Tyr116

His88His90

Glu92

Free energies

D-M 16kcal/mol, T-M 43kcal/mol, pH 7-3.5

D-M 13kcal/mol, T-M 35kcal/mol, pH 7-4

wt1 < wt2 mut_noa < mut_a

Conclusions

Stability, Yes?

Relative stability, No

Important residues identified?

Desolvation, Descreening

Need for a method that allows for

conformational flexibility.

Stability of -crystallins using molecular dynamics

Andy Purkiss

Crystallin X-ray crystal structure to 1.2Å

• A two domain protein, each domain around 80 residues.

• Each domain has a pair of ß-sheets each formed from two greek key motifs.

• Short linker peptide with bend bringing domains together. • A six residue hydrophobic domain interface.

-crystallin features

High Temperature (500K) Simulation of crystallin

0ns

1ns

Wildtype F56A Mutant

Water Insertion Protocol on crystallin

Cycle 1

Cycle 1000

Wildtype F56A Mutant

Cluster Analysis of Water insertion on C-terminal domain of Scrystallin

Cluster Analysis of Water insertion on C-terminal domain of crystallin

• Distribution of the distance of Ile 56 from Helix B and Helix 310

(sitting each on the two opposite sides of the residue) from all the conformations sampled.

• Only in the mutant conformations we find that Ile 56 exhibits two alternative preferable distances from Helix B and from Helix 310 one near the distance of the crystal structure and one distant.

Results: Ile 56 positioning