I'm predicting the protein-peptide structure. Why is the result different everytime?

Question

[soobiney]

Hi. I am a doctoral student.

The binding structure is predicted using specific protein and peptide sequences. By the way, each time I do, I have to worry about how to interpret it because the result is different.

The program used is the web-application CABS-dock. Put the PDB number (only the protein receptor part) in it, put in the specific peptide and wait for 4 ~ 5 hours.

I want to get data from 10 predicted values ​​attached to the desired spot (the middle part of the receptor) and sort out the result when it comes out.

I do not know which value is correct because the result is different every time. I am studying the structural protein field because it is not a specialized field.

I do not understand the meaning of the terminology of CABS-dock. For example, "cluster-density", "RMSD(root-mean-square deviation), average RMSD, max RMSD"

Currently, the structure of the peptide is not known, so I use a prediction program to identify the binding site.

If you have a program that is more accurate and reproducible than the CABS-dock, please let me know.

Thank you.

Answer 1

[Joe]

I'm not familiar with that tool specifically, but the reason you get different numbers each time is likely because the software is modeling some stochastic fluctuations and it isnt an entirely deterministic process. If the software has the option to provide a starting random seed, try making this the same each time you run it.

Ligand docking is a fairly 'wooly' process though, so that all sounds pretty normal to me (in my albeit very limited experience).

To answer your terminology questions:

Cluster density

If this is what I think it is (based on how ITASSER uses this terminology), this is likely how frequently occupied a givel voxel is by the model. I.e. if you run the same model 100 times, and the ligand/protein tend to occupy the same 3D space each time, that's more likely to be the real shape/conformation.

RMSD

RMSD is a way of measuring the Euclidean (straight line) distance between two points, usually atoms (often alpha-carbons) in this context, in 3D space.

Its simply the x, y and z coordinates squared (to remove any sign), added together, and then all square rooted to return the magnitude to the right level. Its defined as:

This can be done for a single atom, or across all elements, (which is where the Sigma comes in).

Max RMSD and average RMSD will be exactly what they sound like.

[Michael]

I think that RMSD needs some clarification to in the context of docking. To calculate it, two atom structures are needed that are somehow supposed to be aligned or similar. It is not clear to me what is actually compared in the docking context, maybe it refers to re-docking and then the software calculates the distance between the experimental docking pose and the predicted docking pose. I'd suggest OP to have a look at the documentation of the software and check which structures are actually are compared.

[Joe]

Yeah that's probably a good point. I assume it would calculate ligand-pocket binding distances (if not all-vs-all), but I'm no expert. a-Carbon to a-Carbon is the default method for protein superimposition studies which I'm much more familiar with.