An improved all-atom force field with implicit solvation for folding simulations of small proteinsEunae Kim and Youngshang PakDepartment of Chemistry, Pusan National University

Our current research2. Folding studies using MD with all-atom level force field with implicit solvation.1. Development of novel simulation schemes

Force field in protein folding Protein force field :united to all-atom models

Solvation model : explicit water and implicit water models

Is the current all-atom force field good enough for protein folding simulation?Not really !

Takao et al, Chem. Phys. Lett. (2004) 386, 460Case1: Comparisons of force fields for Alpha helix C-peptide and Hairpin G-peptide in explicit waterC-peptide using REMUCAG-peptide using MUCAREM* Initial structure : unfolding state

Lesson I

With explicit solvation, alpha/beta imbalances exist.

Pak et al, J. Chem. Phys. (2004) 121, 9184Case 2: REMD simulation of 1PSV using param99MOD2 with the GB implicit water

The current GB solvation model has a problem ! The salt bridge effect is overestimated by the GBSA model

Lesson II

With implicit solvation,

1. alpha/beta imbalances exist. 2. The salt bridge effects are overestimated by the GB solvation.

Our immediate goalFor all-atom force field with GB solvation, native state should be the lowest free energy minimum with reasonable energy barriers.

Proposal : To achieve our immediate goal, adjusting all-atom force field with the GB implicit solvation: Fine tuning these!

The Generalized Born Model Proteins 55, 383 (2004), David A. CaseWhere

In order to weaken the overestimated salt bridge effect by the current GB model, we currently propose as follows:

We employ scaling i of the functional groups of charged side-chains and hydroxyl oxygen by = 0.85.

Force field optimization

Replica Exchange Molecular Dynamics

param99MOD3 The backbone torsional potential parameters should be further modified for more balanced description of motifs (a training set of 1FSD, 1PSV and BBA5)

Formation of the -strand is enhanced by the scheme in / peptides

REMD with our new force fieldInitial structure : Folded structureIn REMD, 20 replicas were run in parallel at the 250 ~ 620 K with TINKER package The temperature gap in REM

Total running time : 50 nsTime step : 2 fsExchange interval : 500 fsTrajectory saving interval : 500 fs

Force field : param99MOD3 Solvation model : the modified GB model (Case et al. & scaling factor, =0.85)Cutoff : 24.0 (the nonbond and GB solvation terms)

Integrate :velocity VerletFriction : 1.0 ps-1SHAKE algorithm

Free energy landscapesBBA51FSD1PSVPak et al, Proteins (2006) 62, 663

Direct folding studies of or strandsEK-peptideC-peptide1e0qgb1Pak et al, Proteins (2006) in press.

Temperature dependence

Direct folding studies of motifs1FSD1PSV

1FSD1PSVThe predicted lowest free energy conformer (blue) NMR structure (gray)RMSD = 2.0 RMSD = 1.8 Pak et al, Biophys. J. submitted.

Problems of param99MOD3 The P state of gb1 is lower than the F state in free energy.(~0.3 kcal/mol)(In explicit solvent, F state < P state : Zhou, PNAS (2003) 100, 13280.)

Incorrect location of the native structure of trp-cage (1l2y: 23-residue protein)

GBOBC II : = 1.0, = 0.8, = 4.85i : intrinsic radii of atom i Table I. Parameter sets used in the various modified GB Implementationsparam99MOD4 (1PSV, 1FSD, BBA5, gb1, 1l2y)

atomRi, GBOBCAMBER8SiRi,modified GBOBCHC1.31.001.3HN (neutral)1.31.001.3HN (basic charged)a1.30.851.105HO0.81.000.8HS0.81.000.8C (neutral)1.71.001.7C (acidic charged) b1.70.851.445N (neutral)1.551.001.55N (basic charged) a1.550.851.3175O (neutral)1.51.001.5O (acidic charged) b1.50.851.275O (hydroxyl group) c1.50.881.320O (Thr)1.51.001.5

param99MOD4

gb1Tm = 22C (Expt. Tm = 24C)

H = 48 kJ/mol , S = 162 J/mol (F -> U)H = 49 kJ/mol , S = 163 J/mol (expt.)

Trp-cage (1l2Y)Tm = 41C (Expt. Tm = 42C)

H = 48 kJ/mol , S = 148 J/mol (F -> U)H = 49 kJ/mol , S = 155 J/mol (expt.)

The simulations for 1FSD, 1PSV, and BBA5 are still underway, but they look promising!

Conclusions Our new force fields (MOD3 and MOD4) show very significant improvement in folding simulation of small proteins:More reasonable free energy landscapes1Correct thermodynamic properties2

Future workUnited atom model (ff03ua) in AMBER9 will be improved using the current scheme.

Free energy mapping and direct folding of more complicated proteins

GB1

Acknowledgments BK21 KISTI

Group membersEunae Kim (Ph. D.)

Changwon Yang (M.S.)

Thanks for attention !!