MD simulation analysis TSD
Contents
General
<figtable id="tbl:generalstats">
| Model | CPU | Runtime | ns/day | days/s |
|---|---|---|---|---|
| Wildtype | 16 | 13h52 | 17.3 | 3896 |
| R178H | 12 | 19h03 | 12.6 | 5354 |
| P182L | 32 | 07h08 | 33.6 | 2006 |
</figtable>
<xr id="tbl:generalstats"/> shows general statistics about the three simulations. It should be noted that the output of gmxcheck does not account for the number of CPUs used in the calculation and only reports the real time that passed. Normalizing the runtimes by the number of CPUs involved yields that the Wildtype and P182L mutation completed within 4 hours. R178H took 4 hours longer, however the difference is negligible, considering that they were not done in a testing environment and it is assumed that all CPUs could maintain equal load throughout the runs. In fact Gromacs reports a particularly high 12% of the time being lost due to particle-particle/Particle-Mesh-Ewald imbalance.
Trajectories
three trajs coming up, need moa ram
well there doesn't seem to be a whole lot happening, ours probaly only makes sense, when you combine two subunit, since only then it's enzymatically active
Thermodynamics
Pressure
<figtable id="tab:pressure">
</figtable>
<xr id="tab:pressure"/> shows the pressure oscillations for the three simulations. As can be seen per-frame values differ by several 100 bar, as to be expected <ref name="gromacsmanualpressure">http://www.gromacs.org/Documentation/Terminology/Pressure</ref>. More importantly the average shows convergence in every simulation and does not undergo any major changes towards the end of the simulations. The final average pressure also lies close to 0 in all cases.
Temperature
<figtable id="tab:temperature">
</figtable>
<xr id="tab:temperature"/> shows the temperature variances during the simulations. The mutations both show higher extreme values than the wildtype structure, especially P182L. The average however remains exactly the same for all simulations, which is the expected behavior after a period of time passed <ref name="">http://www.gromacs.org/Documentation/Terminology/Thermostats</ref>. The fact the all simulations arrive at the same value also supports that the simulations went well and arrived at the correct temperature.
Potential Energy
<figtable id="tab:potentialenergy">
</figtable>
<xr id="tab:potentialenergy"/> shows the fluctuations of potential energy during the simulations. As can be seen, during all simulations it is globally decreasing and the final averages of all three runs are similar.
Total Energy
<figtable id="tab:totenergy">
</figtable>
<xr id="tab:totenergy"/> shows the total energy which is composed of the potential energy shown before and the kinetic energy <ref name="gromacstotenergy">http://www.gromacs.org/Documentation/Terminology/Total_Energy</ref>. It also globally decreases for all runs and the final averages are very similar which leads to the conclusion, that the kinetic energy behaves similarly. Given that the WT behaves in the same way than the mutations one cannot say that the mutations have an effect on the total energy.
Flexibility
RMSF
B-Factors
<figtable id="tab:bfactor_ca_vs_prot">
</figtable>
<xr id="tab:bfactor_ca_vs_prot"/> shows a comparison of the B-factors based on the converted RMSF values. It can be seen that the values calculated only from C-alpha atoms are similar to those calculated from all atoms in the structure. As a result, in further analyses, only flexibility inferred from all atoms will be further discussed.
It can be seen that the mutation residues do not seem to be significantly moving during any of the simulations, whether they are mutated or not. The wildtype shows flexibility in several loop regions at the right of the figures. These loops were also found to be variable during NMA, however no functionality could be linked to them. In addition, E323 shows flexibility based on all atoms and also on the backbone only. This is indeed one of the residues that are essential for functionality as outlined in the introduction.
R178H does not exhibit any flexibility at this position which meets expectations, given that the mutated residue might fit into the binding pocket but cannot coordinate the ligand anymore (c.f. MD preparation). It is interesting to see though, that this behavior is observable in the protein although there was no ligand present during the simulation. The molecule generally appears very rigid, with even the outside loops showing no significant movement anymore.
P182L is not known to cause TSD and analysis from flexibility only also suggests this. All regions flexible in the wildtype remain flexible, however the missing proline, which acted as a helix breaker seems to introduce additional flexibility in the right side of the protein. In the native protein conformation the first domain is located at this side of the protein. It might either give additional stability or be disturbed by the missing rigidity in the contact helix. Which of the two is the case would have to be observed by trying to model the unresolved region in this domain (which is why it was excluded in the first place) and performing additional MD runs. It is noteworthy that one of the residues that gain flexibility, E462, is one of the important residues. Whether the mutation shows biological functionality that is not exhibited in the wildtype or whether, in reality, the mutation is accounted for by the presence of the additional domain, cannot be determined at this point.
Comparison to experimental values
Comparing the RMSF-induced B-Factors evaluated above to the ones present in the original PDB file, not similarities could be observed. The PDB file shows higher overall flexibility and the few residues with highest flexibility are none of those seen before. More importantly the are no residues that could be explained to be flexible by any of the analyses performed so far. Of course the 'real' B-Factor as approximated by the experimental values in the B-Factor is based on more than just movement of the residue. Apparently the RMSF values do not approximate these values very well but instead carry other information, that gives hints to protein functionality.
now check how avg (i.e. rmsf which is per residue mean movement), behaves compared to the b-factor he? isn't the nfactor strcutre based on average? this is probably also comparable? hu? b-factor are converted rmsf values!
wenn nein, prima. dann final nochmal B-factors von prot von den mutationen verglichen mit dem aus dem wildtype. soll heissen, aendern unsere mutationen irgendetwas daran wie sich das ding verhael in terms of flexibility?
look at b-factors in the very original file, that is the experimentally annotated ones. especially how are they in the domain that we cut out?
RMS Fluctuation plots
this is the flucutation of the atom around its average. depending on above it should be ok, if we only show prot! or simply see again manually, whether they look similar
Significance
marcos pi test anschaun see pku
RMSD against average
is that really it?
THESE ARE a different topci!! see CONVERGENCE OF RMSD
in manual! (probably different)
RMSD against start
and probably end as well? check again
Conclusion
References
<references/>
