The Molecular Graphics Laboratory Forum

If anyone could help me to analyze this situation, it would be great...

I've been using Vina to find binding sites of several blocking peptides to a specific membrane-receptor cytoplasmic domain (TIR domain). My first attempt was using a "blind docking" strategy, in which I defined a search-space bigger than this TIR domain. Unfortunately, this approach didn't achieved the results I expected, identifying binding sites that doesn't match with those found empirically.

In order to obtain reliable results, I decided to execute a new docking strategiy called Multi-Grid Strategy (also called Multi-Space Strategy). In this case, the search-space used before is divided in smaller search-spaces, executing Vina afterwards. As you can imagine, this method represents an assembly of many specific docking experiments whose hypothesis are that the correct binding site belongs to the corresponding search-space. After choosing the best binding mode of each docking experiment, an energy criteria is applied to identify the final binding site (the correct binding mode is the one with the lowest binding energy).

Because Vina uses an stochastic docking algorithm, it's necessary to repeat these experiments with different seeds in order to avoid docking results associated with local minima.

In my opinion, this method has been successful because most of the studied blocking peptides reached minimum binding energy values inside search-spaces where empirical binding sites were located. Nevertheless, after analyzing first and second round results, one of these peptides reached different binding energy values and binding sites located in different search-spaces:

1st Round:
(results adapted as Vina's output; non-overlapping search-spaces)

| Search-space | mode | affinity | dist from best mode
| | | (kcal/mol) | rmsd l.b. | rmsd u.b.
+--------------+------+-------------+-----------+----------
A 1 -9.4 0.000 0.000
B 1 -6.0 0.000 0.000

2nd Round:
(results adapted as Vina's output, non-overlapping search-spaces)

| Search-space | mode | affinity | dist from best mode
| | | (kcal/mol) | rmsd l.b. | rmsd u.b.
+--------------+------+-------------+-----------+----------
A 1 -8.4 0.000 0.000
B 1 -9.6 0.000 0.000

I think this is related to local minima or random seed selection, but it seems weird for me such kind of difference between energy values under same ligand and receptor configuration.

Any ideas?

Camilo Patino Salas
Chemical Engineering Student
Universidad de Chile