Temperature REX Simulation of 1E0Q

l1E0Q is Mutant Peptide from the first N-terminal 17 amino-acid of Ubiquitin, check the paper "Structural characterization of a mutant peptide derived from ubiquitin Implications for protein folding" to learn more about this small protein, actually it is a part of a big protein.

The sequence of 1E0Q is:




Temperature Replica Exchange Simulation were performed under vacuum, gb, and explicit conditions. native replica simulation: native structure is used as initial structure of each replica. the command to run REX simulation with MMTSB is:

aarex.pl -n 1000 -par archive,natpdb=1e0q.pdb -temp 8:300:850 -log rex.log -charmmlog charmm.log -elog energy.log -mdpar cmap,blocked,(gb) -dir gbrex_va(gb) 1e0q.pdb 

The simulation performs 1000 exchange, 1000 dynamic steps in between every exchange, dynamic step is 0.002ps. Let's first take a look at the movie of all replicas in vacuum, gb solvent.


To better observe the stucture fluctuation, run rmsd vs time analysis by:The left is vacuum simulation, the right is implicit solvent simulation. As we can see in vacuum simulation the replica are easily spread to random structures while in implicit solvent the replica keeps the more stable at lower temperature, and one helix also show in one replica, this means implicit solvent helps to gain the secondary structure more than in vaccum. 

rmsd_grid.sh gbrex_va

rmsd_grid.png will be generated under directory ./gbrex_va

1e0q_8_va_rmsd_grid 1e0q_8_gb_rmsd_grid


As we can see in vacuum the rmsd at the beginning were very low about 2~3A, after about 100 cycle, they all lost their native structures, but in implicit solvent, the first 3 replica have very low rmsd, and did not exchange to extended conformation. To check the energy overlap of all the replicas, run command:

ener_overlap.sh -byclient(-bytemp) gbrex_va

energy_overlap_replica.png or ener_overlap_temp will be generated under ./gbrex_va.


1e0q replica in vacuum                                                                                                                     1e0q temperature in vacuum


1e0q replica implicit                                                                                                                               1e0q temperature implicit

Each color represents one replica(byreplica) or temperature(bytemp),as we can see the energy of each replica spans almost all energy range. The distribution of each temperature follows boltzman distribution,and they are overlaped about 1%~5% which is less than we expected(30%~60%), this is due to the large range of temperature and less number of replicas, later we will show the data of simulations with more replicas. 

The exchange percentage from  rexserver.ninx are around 2%~5%. To get higher exchange rate, we need to increase the number of replicas, follwing are the energy overlap of 16 and 32 replicas.


The exchange rate of 16 replicas are around 25~30%,the exchange rate of 32 replicas are around 55~70%. Let's look at the rmsd for explicit solvent for 32 replicas. 1e0q_32_ex_rmsd_grid The exchange rates of explicit solvent are 0%~17%. This is due to the calculation of energy, because the total energy include the energy of waters, thus make the exchange harder to happen. As some papers pointed out, replica exchange was not born to do explicit solvent simulation. 

Conclusion: for 1e0q which has 17 residues, 16 or more (24,32) replicas in implicit solvent is good enough, do not waste time to run explicit simulation.

The above results are from REX simulation of native structure, following are the results from random conformations in vacuum,gb,and explicit solvent.These are from 24 replicas.




The top movie is from vacuum, and bottom one is from implicit, the bottom one has more helix than top one, this suggests the implicit solvent model is preferable to the formation of helix. None of them show the formation of hairpin.


The left side is vacuum, and the right side is implicit, it seems implicit solvent model helps in the exchange of replicas, since the color change more dramatically than vacuum, but if we look at rexserver.ninx, the exchange acceptance fraction are about the same value around 50%. From the movie we can tell in implicit solvent the structure fluctuate more dramatically than vacuum.