The "RMSD" of the energy is the standard deviation. In energy minimization, it is completely meaningless because you do not expect energy conservation; the purpose of EM is to change the structure and relax its energy. This often changes the potential energy by many orders of magnitude.

Hi,

The general criteria of significant RMSD is less than or equal to 2Å.

You can consider following articles for detailed information:

https://pdb101.rcsb.org/learn/guide-to-understanding-pdb-data/resolution

https://www.researchgate.net/deref/https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fpmc%2Farticles%2FPMC3057020%2F

The "RMSD" of the energy is the standard deviation. In energy minimization, it is completely meaningless because you do not expect energy conservation; the purpose of EM is to change the structure and relax its energy. This often changes the potential energy by many orders of magnitude.

thank you for your responses Fareha Masood and Justin Lemkul .

Dear Ehsan Alirahimi ,

EM is a procedure to make the structure compatible with the Classical-Mechanics forcefield. Every current forcefield has a bias and the final structure after EM will deviate from the experimentally determined structure, where QM is the dominant player.

The RMSD only tells you how far the structure is departed form the starting point, and it should be as small as possible, assuming the starting structure was right. The most important aspect is the final energy, which should be large in magnitude and negative in sign. Your plot have no axis or units, so it is uninformative. But the screen shot below shows a low energy, and that is what you want.

Dear Rogelio Rodríguez-Sotres

thank you for your response,

so the high score in the RMSD field that i got can be explained by two possibilities: a)either the model that i obtained after EM is very much different (in structure) in comparison with the experimental derived structure, b) the experimentally derived structure was not of high quality.

considering you second point (final energy score) how large is large enough to be happy with your EM results (i mean is there any threshold?).

and lastly , i have attached my EM c plot to this email and i really appreciate your comment on that.

all the best.

If I may offer some tutorial material of interest:

http://www.mdtutorials.com/gmx/lysozyme/index.html

as well as:

https://www.livecomsjournal.org/article/5068-from-proteins-to-perturbed-hamiltonians-a-suite-of-tutorials-for-the-gromacs-2018-molecular-simulation-package-article-v1-0

The magnitude of the potential energy is an extrinsic property and is largely based on how many water molecules there are. Your structure will indeed deviate from the starting structure, due to force field effects but more importantly due to two other considerations: (1) crystal structures are not the same as solution structures and therefore you should expect change and (2) even NMR structures are not literal structures; they are interpretations of data that may or may not be entirely accurate. Many structures are underdetermined. Of course, the force field is the biggest influence.

The threshold for termination of energy minimization is the maximum allowable force that you set (the gradient), not the potential energy. In single precision, you may not be able to achieve any arbitrarily low value of the gradient, but it usually does not have a huge impact on the simulation as long as a sufficiently low value is achieve (see tutorial material linked above).

To reiterate my point above, the "RMSD" that you see has nothing to do with the structure of the biomolecule. It is the standard deviation of the energy series, which in this case has no physical relevance. You also cannot use structural RMSD in any productive way here.

Dear Ehsan Alirahimi ,

take Justin Lemkul advice, follow the tutorial, it does not hurt.

Now regarding your plot:

The plot you included is Energy (kJ/mol) vs time (I guess is more "frame number" that time, because time is not implicated in EM).

All it means is the energy "as calculated with the forcefield" has gone down from roughly -25000 to approx. -42000 kJ/mol.

That is exactly the aim of EM, to drop the energy down, by moving atoms little by little, following the energy gradient as a function of atom coordinates, for the whole system.

What does it means for your protein?

A) the Experimental structure is what we know about the molecule from sampling the real stuff.

B) Your forcefield is a good approximation to the calculation of energies for real systems, but it is an approximation anyway, not perfect. It does consider the average bond distance for bonds of C-C, C-N, C-H, C-O, etc. the average bond angles for C-C-C, C-C-N, C-C-H, etc. the dihedral torsions for single bonds and for atoms groups, i.e tetrahedral geometries, planar bonds, etc. All that is approximated using classical Newtonian mechanics, without introducing QM electronic effects, because consider them would make the calculation too costly to be of practical use.

C) To make your system completely consistent with the forcefield you need to adjust all the values mentioned in B, not to the "true real-life" value, but close to the "optimal forcefield value". That is what geometry optimization does.

You want to see how much your protein has changes? take a molecular visualization tool (VMD, CHIMERA, YASARA, PYMOL, etc) load the starting structure and the minimized structure, align them and calculate RMSD (should be less than 2A if everything went well), but more important, analyse the structures and find is anything in there looks suspicious.

If you are using gromacs, you may also use:

gmx rms -s em_input.tpr -f em_trajectory.trr -o rmsd.xvg

"rmsd.xvg" can be plotted with xmgrace (or grace) to show how the rmsd varied along the run.

best wishes,

Rogelio

Hi, The RMSD is the standard deviation of the energy from its original value - i.e. the y-axis of the plot has the units of the gromacs energy (E (kJ/mol)), which has nothing to do with the RMSD to the native structure. Indeed, it would be perfect to minimize the free energy within 500 steps - but that is impossible yet. The x-axis title is not Time (ps), but the number of minimization steps (# N_step). Your structure seems to be well optimized for its potential energy, so you could start to run the MD simulation.

Dear Justin Lemkul

thank you for your response. that was a great help. I also had a look at the article that you provided the link of and I found it very useful. I will read it thoroughly.

Best,

Dear Rogelio Rodríguez-Sotres

thanks for the information.

I learnt many great things from your explanations.

Best

Dear Emanuel Karl Peter

thanks for your answer.

in the potential.xvg file that was generated, it is written that:

.

.

.

@ title "Gromacs Energies"

@ xaxis label "Time (ps)"

@ yaxis label "(kJ/mol)"

.

.

and then i copied and pasted the numbers in Excel and made the plot. your are in fcat right about it. it should be the number of steps (i remember the last step in EM was numbered as 690= the last number in the potential.xvg file).

I have no explanation for it.

any idea?

thanks again,

Best

Just a clarification:

RMSD = "root mean squared deviation"

Is is calculated by identifying the "equivalent" atoms on two molecules, then all pairs of corresponding atoms are aligned to reach the lowest possible difference in their positions (without distorting the structure). Then the euclidean distance between them is calculated and squared, the squares are added up, divided by the number of atoms, and the square root of the result is calculated.

There is no involvement of energy at any point, its dimensions are LENGTH, i.e. its more common units are Å or nm. Two identical structures have an RMSD of ZERO and two completely divergent structures have an RMSD of about 1/2 of their average molecular stokes radius.

You can read more about it in:

Article Methods of Protein Structure Comparison

Best wishes,

Rogelio

= 1/N_t sum_i^N sqrt(x_0 - x_i)^2 can be applied to any kind of property.

= 1/N_t sum_i^N sqrt(E_0 - E_i)^2 is the case for energetic quantities over time or in that case the minimization steps. The output of g_energy is linked to energetic properties. Please let's stop here. It is getting serious.

Sorry but this need to be said:

I agree the calculation may be applied to any value. But by definition RMSD is applied to distance, or sometimes to angles. If you apply it to energy, it should be indicated.

The plot above seems plain energy, unless you took a rather distant and external reference, and not the first frame of your run, which should make the plot start at zero and grow, because the square root value taken is the positive one by definition.

The equation by the way is:

RMSD = SQRT( (1/N) · SUM_i=1^N (||v_i -u_i||^2)) )

where u and v are two sets of point with bi-univocal correspondence.

if you take the SQRT of the difference and then square it before doing the sum, you get the sum of absolute values, not the sum of squares.

It is all in the paper I quoted

Best wishes

Rogelio

Dear Emanuel Karl Peter ,

thanks for your answer.

Dear Rogelio Rodríguez-Sotres ,

Thank you for your explanations. I appreciate it.