How MoveableType solves three challenges in binding affinity prediction in drug discovery

Successful drug discovery relies on the free energy of binding (i.e., binding affinity) predictions, as it is the single most important initial indicator of drug potency. Computational molecular simulation protocols such as docking and scoring, Monte Carlo (MC) and molecular dynamics (MD), support the need for large-scale virtual screening during the hit-to-lead stage of drug discovery, helping to limit the trial-and-error cost of designing and synthesizing drug leads. However, it is challenging to accurately, reproducibly, and quickly predict binding affinity and protein-ligand binding modes with today’s computational molecular modeling technology.

Fortunately, MoveableType (MT), a game-changing computational tool, is the solution to overcome three challenges of conventional molecular modeling software.

MoveableType – a free energy-based method for characterizing binding modes, conformational distributions, and absolute binding affinity. It’s fast, accurate and easy-to-use.

Challenge 1: Optimizing molecular simulations for accuracy and reproducibility

Typically, docking and scoring methods capture global minimum binding modes using pose-generation plus restricted local sampling followed by protein-ligand binding free energy estimation using a single “winning pose” (as chosen using a selected score function)..

Using only local pose-sampling and single binding mode conformations threatens the accuracy of binding affinity predictions as the models do not always capture the fact that many proteins change conformations depending on their environment, activity, and bound state (i.e., unbound versus bound ligand state). This approach can also lead to biases where larger molecules outperform smaller ones because there are simply more intermolecular pairwise contacts.
Having more information on the different conformations for both bound-state and free-state protein-ligand pairs is critical for accurately predicting binding affinity.

MovableType maximizes molecular sampling by relying on protein:ligand ensemble conformations

MovableType enables more comprehensive surface sampling as it relies on ensemble sampling to successfully predict high-affinity compounds and non-binders. This approach incorporates multiple protein:ligand conformations to maximize molecular sampling space. Each protein:ligand conformation is considered a representative landscape minimum, and when combined, this ensemble captures “global” molecular movements and provides a more detailed representation of the protein structure and how it changes.

MovableType enables flexible sampling approaches to optimize molecular modeling simulations

The primary MT protocol is to couple it with docking modes generated with fixed-protein:flexible-ligand docking or induced-fit (or flexible-protein:flexible-ligand) docking. These modes are minimized, curated & treated as landscape minima. While MT coupled with docking works very well for many cases, occasionally further optimization and increased sampling is needed.

With recent protocol and theory improvements, MT is now able to calculate the absolute binding affinity based not only on docked poses but dynamics snapshots, hand-modified models, or even multiple experimental targets/poses. These improvements expand the domain applicability of the MT method to protein targets, which require additional protein:ligand sampling beyond what is available from docking alone.

QuantumBio has even developed four stages of sampling progression for these particularly complex cases.

MoveableType performance with different sampling methods

The MT free energy method uses bound-state and free-state conformations of protein-ligand molecular systems to estimate absolute binding free energies.

We used the Merck KGaA validation sets to explore the compatibility of the MT free energy method working with different sampling methods, and show that the method is not only applicable as a virtual screening algorithm by using docking-generated poses, but also as a rigorous free energy evaluation algorithm by using protein-ligand complex trajectories from molecular dynamics simulations.

Merck KGaA Benchmark Results Analysis
Pearson’s R: comparison of the different free energy protocols
*Those highlighted in blue have R > 0.5

Protein No. Ligs IDF Dock + MT-amber IDF Dock + MT-garf cMD + MT-amber cMD + MT-garf
CDK8 33 0.407 0.477 0.463 0.552
c-Met 24 0.724 0.822 0.749 0.852
EG5 28 0.483 0.394 0.582 0.584
EG5 (alt) 28 0.504 0.505 0.494 0.542
HIF-2a 42 0.107 0.237 0.257 0.507
PFKFB3 40 0.101 0.056 0.689 0.840
SHP-2 26 -0.065 0.156 0.232 0.416
SKY 44 0.204 0.300 0.321 0.411
TNKS2 27 0.591 0.541 0.705 0.640

IDF: Induced Fit
cMD: conventional Molecular Dynamics
MT-amber: MovableType with AMBER force field
MT-garf: MovableType with GARF force field

The present work shows a mixture of cases in which docking paired with MT is sufficient (CDK8, c-MET & TNKS2) and cases that require further optimization via additional sampling available in MD+MT (EG5, HIF-2α, PFKFB3, SHP-2 & SYK).

Learn more by reading the full publication.

Challenge 2: Overcoming the speed versus accuracy tradeoff

Speed versus accuracy has been a much-talked-about dilemma when it comes to computational molecular modeling. Docking-based approaches are fast, yet lack reproducibility and reliability. Molecular dynamics-based approaches are more accurate but can take days or weeks to know whether the simulation failed. Drug discovery teams need computational tools that are both accurate and fast to minimize the hit-to-lead candidate timeline and push clinical candidates likely for success through to synthesis and experimentation.

MovableType is fast and maintains accuracy by combining molecular docking with localized and global sampling

MovableType has shown itself to be competitive with other, much more computationally expensive methods. While MovableType can utilize molecular dynamics snapshots to represent protein:ligand models, docking can also be used to produce these models – this approach is often quite predictable. Therefore, MT is often able to achieve much higher throughput compared with lengthy molecular dynamics and conventional free energy-based methods, but with similar accuracy.

By enabling rapid simulations, researchers can observe the results in near real-time, allowing them to decide if they should synthesize or abandon a structure as early as possible.

Method Sampling Method Pose Generation      Average Time
(CPU min/protein-ligand complex)
MTScoreE MOE induced-fit docking Docking prose and generation ~13 min.
Conventional MD + MTScoreE 250ns cMD simulations using Amber18 with ff99SB force field Complex pose generation ~1180 min

Challenge 3: Making molecular simulations convenient and user-friendly for drug discovery teams

Conventional molecular simulation software is very complex and includes many settings and features to tweak and specific hardware requirements (e.g., supercomputer, GPU) must be satisfied to complete these calculations in a timely fashion. This can make it difficult to bring into the industrial/commercial pharmaceutical lab and sometimes even more challenging to get scientists motivated to use it.

MovableType is less complicated, easier to use, and simple to integrate

In the default protocol, there are no molecular dynamics simulations to set up and manage, and no complicated or exotic hardware or software required. MT offers consistent execution and convergence using standard input formats and fewer settings to tweak.

Furthermore, MovableType is an absolute binding free energy (∆G) method, while Free Energy Perturbation (FEP) is generally a relative binding free energy calculation (i.e., ∆∆G). Absolute calculations are sometimes less accurate for systems with minor local structural differences. However, MT balances this with ensemble-based sampling and multiple protein:ligand sampling options. Together these features increase software accessibility, especially for high-content screenings with multiple targets, so even non-experts can run simulations quickly.

And MT is designed to be as efficient as possible for IT professionals to get it up and running. The command line, graphical user interface (through MOE/svl), and web service can all be integrated with JSON API.

MovableType is flexible in the way it is used and what it is used on

With pharmaceutical companies becoming more and more collaborative and rapidly changing, we realized that our software needs to be compatible with changing applications, software ecosystems, and user’s screen of choice.

Here are the different ways you can use MovableType:

  • On premises: It can run on conventional clusters, workstations, and even laptops with ease.
  • Cloud access: It can be run on GridMarkets Pharma’s cloud-based platform.
  • 3rd party dockers: Researchers can use a built-in heat map docker or a 3rd party docker (e.g. MOEdock) through SDF support or with molecular dynamics by using PDB snapshots.
  • One of many simulation protocols: Depending on your application, you can choose between various simulation protocols for rigid-receptor docking, induced-fit docking, MD/MC snapshots, cross-docking, etcetera applied to both apo and holo target models.

Modern molecular simulations software is the keystone to reduced-cost, accelerated therapeutic pipelines

Computational molecular modeling for prediction binding affinity has changed the game in drug discovery, as it helps guide the decision of which compounds to chemically synthesize for wet-lab experiments. However, conventional software has several limitations that extend hit-to-lead candidate timelines and result in costly drug development activities. To identify drug candidates that are most likely to succeed, minimize expenditures, and accelerate therapeutic pipelines, pharmaceutical companies need modern molecular simulation software solutions like QuantumBio’s MovableType.

To break through the conventional time and cost barriers of in silico drug discovery molecular modeling, MoveableType empowers scientists with five innovative key features:

  1. Accuracy: MT calculates binding modes, conformations, and affinity on the energy surface by utilizing localized and “global” sampling coupled with a pairwise potential to capture the dynamic nature of protein targets.
  2. Fast and High-throughput: MT performs molecular simulations in CPU-minutes rather than CPU-days, without the need of expensive supercomputers, specialized hardware, or even graphical processing units (GPUs).
  3. Flexibility: Use the software on your choice of screen and adapt the protocol based on your needs for rigid-receptor docking, Induced-fit docking, dynamics snapshots, and apo and holo target models.
  4. User-friendly: In the default, docking-based protocol, there is no molecular dynamics to setup and manage, no complicated or exotic hardware/software required. It offers consistent execution and convergence using standard input formats and fewer settings to tweak.
  5. Integratable: Implement the command line, graphical user interface (through MOE/svl), and web service with JSON API.

Transform your drug discovery workflow for the better with QuantumBio’s software solutions.

Learn More About MoveableType


QuantumBio offers a powerful suite of innovative software products purpose-built for life sciences on cutting-edge science that utilizes the highest levels of theory available to achieve peak accuracy, performance, and versatility. Through our science-first, customer-centric approach, we make precision quantum mechanical approaches more user-friendly, cost-effective and easily accessible. We help pharmaceutical, biotech, and academic scientists improve their understanding of biochemical structure and function while enhancing the drug discovery process.