The term of art we are looking for is “landscape minima.” The critical component is not docking per se, but the accuracy and applicability of the landscape minima we are supplying to the MovableType (MT) calculation. The more accurate/representative the landscape minima provided, the better our predictions. 

If we have an experimental (X-ray, NMR, or CryoEM) structure, it can be used as a landscape minimum and, assuming the experimental structure is accurate, it is possible that no further poses are required. However, when we are working with novel compounds which have not yet been made in a test tube, we need a computational method to mimic or represent  the experimental structure(s) for the ligand. 

We can use molecular dynamics (MD) or docking or even hand placement. Of course MD and hand placement tend to take the most time making them impractical for higher throughput screening. Therefore, our default approach is to pair MT with a docker of some sort. In our experience, your best chance at success with MT is to let the docking software you choose spend the time it needs to generate its best landscape minima instead of trying to make docking as fast as possible. MT—when compared to other free energy methods—will more than make up for the extra docking time.

In our experience, MOE (with our supplied/SVL settings) is the most compatible 3rd party docker available, but MTScoreE has been successfully used "in the field" with MOE, GLIDE, GOLD, OpenEye, DOCK, and other docking functions, and the the method has been implemented to be docker agnostic.

You should choose the right docker for your experiment and you should take particular note of the docker-specific settings required to maximize predictability (e.g. be sure in situ optimization of each pose is turned ON and if possible, use induced fit docking—which modifies the target as well as the ligand—instead of rigid-target docking). Read more on docker compatibility.

There are critical benefits of MTScoreE (E = Ensemble) scoring versus other, more conventional scoring methods. Our method takes into account numerous conformers, as opposed to one single pose, and uses a little (or a lot) of information from each pose using what’s called a boltzmann average.

Conventional scoring functions rely on the user providing a single, accurate, active protein:ligand complex pose in order to calculate the binding affinity. The keyword there is accurate. If it is inaccurate, the scoring function will likely be wrong as well. It is difficult, and often expensive, to obtain the right pose (in fact, usually the score function is used to determine which score function is correct).

“In the test tube” or “in the body”, true binding affinity between a protein and its ligand is dependent on many poses--an ensemble of poses. Usually the ligand doesn’t bind a single conformation. And often these conformations are similar to each other, but sometimes they can be quite different. By accepting a series of conformations or poses, MT is able to bring together the information contained in the set and calculate a binding free energy.

For more information

• How does MTScoreE w/docked poses compare to MTScoreES w/X-ray poses? See: MovableType: Frequently Asked Questions
• How important are quality docked poses to the MT process? See: MovableType: Frequently Asked Questions