I am in search of the exact procedure for autodock result analysis. The results are generated in a way that is somewhat complex to understand.
Some of the software, like Discovery studio (Ligandfit) and others, provide score like manners, but in autodock it is confusing for me. Please help me to analyse the pose and get the results.
Dear Vishal,
Mostly Docking programs uses scoring functions for probable binding affinity between protein and ligands. For instance;
If you are using 0Auto dock vina then you will get RMSD values for each ligands interaction. Lowest binding energy shows best affinity with respective 0 RMSD value. For more detail; go through this article
(http://www.ncbi.nlm.nih.gov/pubmed/19499576)
Moreover, Docking results can be visualized by PyMol in order to get atomic binding distance, bonds and what kind of amino acid or nucleotides are involved between protein and ligands.
good luck
Have you ever tried Autodock Vina instead of Autodock? The docking routine is improved in there, and the data output + docked structures is very intuitive.
Please, illustrate me about docking. I read some months ago a paper showing that using several available docking programs the results become kind of random number generator. I'm not sure if there is some program which provides systematic coherent results for several problems... I would like to see how several probes dock to the same protein and also how the same probe docks to several proteins. Is such an study published. If not: boys start doing it!
open .dlg file in text pad and search for RMSD TABLE in that
Like shown below
RMSD TABLE
Rank |Sub- Rank|Run|Binding-Energy|Cluster-RMSD|Reference-RMSD|Grep-Pattern
1 1 45 -6.24 0.00 83.60 RANKING
1 2 25 -5.50 1.11 83.60 RANKING
1 3 20 -5.36 1.53 84.30 RANKING
2 1 4 -4.03 0.00 90.64 RANKING
2 2 23 -4.03 0.26 90.62 RANKING
2 3 2 -4.00 0.21 90.62 RANKING
2 4 42 -3.98 0.38 90.70 RANKING
2 5 33 -3.96 0.23 90.60 RANKING
2 6 27 -3.96 0.26 90.45 RANKING
2 7 18 -3.96 0.25 90.69 RANKING
2 8 12 -3.96 0.41 90.59 RANKING
2 9 31 -3.94 0.24 90.48 RANKING
2 10 47 -3.94 0.63 90.38 RANKING
2 11 5 -3.93 0.64 90.62 RANKING
2 12 28 -3.93 0.19 90.54 RANKING
2 13 10 -3.93 0.53 90.68 RANKING
2 14 48 -3.93 0.54 90.72 RANKING
2 15 26 -3.92 0.50 90.74 RANKING
2 16 13 -3.92 0.61 90.63 RANKING
2 17 15 -3.91 0.54 90.61 RANKING
2 18 29 -3.91 0.62 90.67 RANKING
2 19 17 -3.91 0.40 90.64 RANKING
2 20 43 -3.90 0.70 90.62 RANKING
2 21 35 -3.89 0.38 90.53 RANKING
2 22 36 -3.89 0.81 90.57 RANKING
2 23 24 -3.89 0.35 90.66 RANKING
2 24 38 -3.88 0.61 90.60 RANKING
2 25 44 -3.87 0.59 90.63 RANKING
2 26 46 -3.86 0.50 90.86 RANKING
2 27 50 -3.86 0.43 90.62 RANKING
2 28 1 -3.85 0.61 90.67 RANKING
2 29 37 -3.83 0.56 90.77 RANKING
2 30 39 -3.80 0.47 90.78 RANKING
2 31 21 -3.78 1.62 89.59 RANKING
2 32 34 -3.77 1.59 89.60 RANKING
2 33 11 -3.75 0.56 90.82 RANKING
2 34 40 -3.74 1.64 89.68 RANKING
2 35 6 -3.74 1.45 89.77 RANKING
2 36 49 -3.64 0.78 90.68 RANKING
3 1 30 -3.81 0.00 89.88 RANKING
3 2 14 -3.71 0.25 89.90 RANKING
3 3 9 -3.69 0.20 89.90 RANKING
4 1 16 -3.69 0.00 92.38 RANKING
4 2 8 -3.65 1.84 91.88 RANKING
4 3 3 -3.58 0.84 92.64 RANKING
4 4 7 -3.52 0.83 92.51 RANKING
4 5 41 -3.52 1.00 92.72 RANKING
4 6 19 -3.50 0.99 92.65 RANKING
4 7 32 -3.47 1.15 92.82 RANKING
4 8 22 -3.39 0.70 92.53 RANKING
In this particular dlg file 45th run has low binding energy and u can visualized whether molecule in site the binding site or not
agree about autodock vina use instead of autodock. The developers themselves underlind better performance of former.
Concerning evaluation of binding mode I prefer to rely on visual analysis and reproducibility rather than on ranking.
First, are you using Autodock 4.0 or autodock vina, because they're both completely different in terms of speed, execution and analytics. Assuming you are using the latter, Autodock Vina, it¡'s the simplest software for docking because the output you see in the --log option (also printed above) is telling you the pose number, the energy (the more negative the value, the better the interaction, which is assumed to be the dG), and then, RMSD's which are useless as they only tell you the structural difference between your first pose and the following poses when compared to the best one.
Ideally, one should trust that the first pose is always the best one (this is a milestone), and so, you must consider if your correct pose is the 2nd or the 3rd. How many poses you need to consider in your analysis ?. Well, is another tricky question, but can be done using X-ray known inhibitors and re-dock them to your binding site and try to see if the software successfully reproduces the binding pose. If your x-ray known inhibitor matches with the first pose, then you only consider the first pose of every molecule. If your x-ray inhibitor matches with your 2nd pose, then, consider the first two poses of every molecule and so on.When doing that, you filter out a lot of raw data you won't need to process, and only focus on those poses where realistically, you may have a computational hit.
Afterwards, would be very recommended to re-score the filtered ligands with another software, preferentially with a force-field dependent scoring function. There are a lot of force-fields to be used (MMFF94, PFROSST...), what to use, is up to you. When applying a Force-Field scoring function, the energy value of dG of some pre-selected ligand will drop down very dramatically, then, you can reject those compounds if you want. Only those compounds where AutoDock vina and another scoring functions gives you good values, should be considered for further analysis.
Additionally, you don't have to trust completely the docking prediction. You have to evaluate if the generated docked poses are realistic or not as Viachaslau Bernat pointed out (i.e, there are no positively or negatively charged groups in the nearby of neutral or aliphatic regions, if there is complementarity between your predicted poses and your binding site, if they poses you obtain are encompassing a known pharmacophoric point described in the literature, etc.). The protocol itself is an art, can NOT be extensively applied for every protein.
If you have more computational resources, those poses of those ligand that look promising can be included in a higher-computer demanding approach, such is molecular dynamics, but this is another issue.
As Mr. Moral has said the more negative the values are the best is the interaction. But, always go in for the best score of the largest cluster. B'coz the largest cluster indicate the suitable docking poses of the ligand. You are also provided with the inhibition constant that can be further used for validation on comparing with the experimental data.
As Mr. Moral has said that re docking of the inhibitor in X-ray data and the analyzing the run with binding energy and H bonding of amino acid of each pose the select the pose which show the same result which is for the original one. use this as a filter for your new ligands.
Sorry! But nobody has answered my previous questions... except perhaps Jesus Seco two days ago who said: ' You don't have to trust completely the docking prediction'! Then my mind can easily say: If you cannot trust the prediction: why you keep using such untrustable programs? Another question I will be happy if it can be answered by anyone in any way tou wish...
As any simulation program, docking software only tries to fit the molecule within the active site without considering any external factor that could dramatically affect to the binding process (presence of structural waters, protein mobility, presence or other co-factors, etc.). That's the reason you don't have to trust the docking results except if you have an experimental validation. The reason everybody uses docking software is tricky, probably because is the easiest way to start when performing any drug discovery project, otherwise, the computational efforts goes up dramatically (imagine have to run molecular dynamics simulations for every ligand you want to consider in every pose the docking software predicts, that's senseless!!! and crazy!!).
About the random number, is similar to what happens with molecular dynamics, where a random number determines the initial velocities of atoms. Here, the random number determines how to start the docking software starts to place the molecule within the binding site. This is not a major limitation because you are not running only one iteration, you are supposed to run as many iterations or runs or whatever they call as you want. Ideally, the larger the iterations, the better the prediction, but also, the longer the time to wait to get some results. There is no consensus, and every software recommends a different number of iterations, but with 50 or so you are carrying out exhaustive search about how to place the ligand in the protein binding site. Moreover, some docking software are "exhaustive", which means the docking does not depend on a specific and predefined number of runs (i.e, it places the ligand in as many positions as the algorithm determines, and then, clusters the results and then prints out the results), thus, no random effect is occurring. As example, Autodock Vina, which for me is the best one (also because is free) uses a random number but you can specify a parameters called "exhaustiveness" to increase the quality of results. However, regardless of the different combinations of random number and exhaustiveness numbers you use, the prediction is essentially the same (or some minor structural differences could be noticed, but the binding mode is practically identical, you might try this as practical and ilustrative example).
About the other question you faced (atom probes used in docking), I don't recommend to tackle this problem through docking, because for doing so there is another software called GRID, which is based on Molecular Interaction Fields (MIF), which is pretty good. Another possibility would be to perform molecular dynamics with binary solvent (if you want, I recommend the paper I wrote "Binding site detection and druggability index from first principles", J.Med.Chem, and if you read carefully (also some external references herein) you will realize how you can perform atom probe through protein surface in different ways (by the way, this is called computational solvent-mapping).
I hope I answered everything. If not, feel free to ask again.
Cheers!!
Whaw! Now I'm a bit better informed. Still you have not answered in full my question.
But anyway, that's fine for the moment...
I'm not an expert in the topic but I know that different codes (i.e. different algorithms) may give quite different docking predictions. In some cases, the predictions are far from being perfect if you allow the molecules to relax which is not necessarily considered in docking software.
@jesus seco moral
Can you recommend any software preferably free with a force-field dependent scoring function to re-score the filtered ligands.
As far as i know crystal structure is just a snap shot of the protein-ligand complex. Though the comparison with interaction present between protein and ligand in crystal will increase the reliability of result BUT will it not eliminate ligand with unique poses.
Ligplot is working best. But it only allows single ligand receptor. If we hvae large set of pdb files of docked poses than it is impossible to screen that.
Similarly, iGEMDOCK provide effective analysis platform, But we can import data from VIna or not. i am not sure regarding that.
Dear Manikandan Alagumuthu
Please brief how to use the post docking process with Autodock Vina results. Actually it provides out.pdbqt file. Please suggest some ideas for viewing this results. Thank you
If the ligand_out.pdbqt file has multiple poses--each as a different model--you can step through the models using the arrow keys in the lower right-hand corner of the PyMOL GUI.
@Ganesh, open pyRx and load the out puts from the .mgltool folder, then you convert the pdbqt to pdb and save it. Open pymol and then launch the above file to visualize the protein-ligand complex.
Try this: http://autodock.scripps.edu/faqs-help/how-to/how-to-setup-adt-scripts
In Utilities 24 there are scripts for analysis.
Take the best conformation and save it to a .pdb format, then use Discovery Studio to obtain data like interactions.
Go to autodock tools folder-Utilities24 folder, and then look for the following scripts: write_lowest_energy_ligand.py, compute_AutoDock41_score.py and summarize_results.py.
how can I interpret autodock (not autodock vina) using discovery studio. what all should be considered as a result? what about their cutoffs to consider as a good dock
Do I take the best conformation from the rmsd table@Angel De Jesus Mendez
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Enzymology
- Enzymes