How to Use DynamicBind
Commercially Available Online Web Server
DynamicBind guide to induced-fit protein–ligand docking, apo receptor preparation, sampling, cLDDT, and affinity-score interpretation.
DynamicBind predicts ligand-specific protein–ligand complex structures starting from a receptor structure and a small-molecule definition. It is designed to accommodate receptor rearrangement during docking, making it relevant to apo structures, predicted proteins, and pockets that may change conformation when a ligand binds.
On Neurosnap, provide Input Receptor and one Input Ligand, choose the sampling and refinement settings, and compare the returned complexes using pose confidence and affinity-style scores. Applications include proposing a binding mode when no holo structure is available, exploring whether a pocket can accommodate a ligand, and comparing ligand-associated receptor conformations.
DynamicBind versus fixed-receptor docking
GNINA offers a docking workflow with an explicit search box and CNN-based scoring. DynamicBind instead jointly proposes ligand placement and receptor changes through a learned generative process. It is useful when a fixed receptor may be an inadequate starting assumption, but a generated pocket rearrangement still needs structural and experimental scrutiny.
The word “dynamic” does not make the output a molecular dynamics trajectory. Reverse-diffusion updates are inference steps, not physical time, and generated motions do not determine binding pathways, rates, or conformational populations.
How DynamicBind Works
How induced-fit complex prediction works
DynamicBind uses an equivariant generative model to change the receptor–ligand configuration while building a proposed complex. This can produce a ligand-specific receptor conformation instead of requiring the supplied protein to already resemble the bound state. See the official DynamicBind documentation for the underlying method and examples.
The modeled flexibility is useful for testing an induced-fit hypothesis, but it does not establish that a large loop movement, cryptic pocket, or alternate conformation is energetically accessible in the real protein. Compare the final structure with the input and with known structural constraints.
Prepare the receptor and ligand
Upload the relevant protein through Input Receptor. Inspect chain continuity, important side chains, missing loops, termini, and confidence in a predicted pocket. Keep enough context to evaluate the biological surface. A poorly defined input region should not become trusted merely because a docking model produces a compact result there.
The Neurosnap receptor input removes nucleic acids, heteroatoms, and waters. Consider whether a metal, cofactor, glycan, structural water, or nucleic-acid partner is essential to the pocket. Their presence in the original structure does not mean they remain part of the modeled receptor environment.
Provide a single molecule through Input Ligand, checking stereochemistry, connectivity, formal charge, and protonation. When starting from a PDB ligand or small peptide, use an appropriate molecular conversion workflow and verify the resulting chemistry; coordinate-only files can omit bond-order information.
Keep receptor preparation and ligand chemical state consistent across a comparison. Changes to construct boundaries or protonation can alter the question as much as changing the compound itself.
Select samples, inference steps, and model settings
| Setting | Practical use |
|---|---|
| Number Samples | Number of proposed complexes. The default is 10; additional samples broaden exploration and increase computation. |
| Inference Steps | Coordinate-update steps in generation. The default is 20; these are not MD timesteps. |
| Model Version | Version 2 is the first available choice; Version 1 corresponds to the original paper model. Record the selected model for comparisons. |
| Relax Structure | Applies post-prediction relaxation. Inspect whether it removes local clashes or changes the pocket substantially. |
| Noise Structure | Controls final-step stochastic noise; keep it consistent for a controlled comparison. |
| Random Seed | A nonzero value records deliberate stochastic sampling; 0 requests a generated seed. |
Start with the defaults, then change one setting for a specific reason. More samples cannot repair incorrect ligand chemistry or restore an essential cofactor omitted during preparation. A seed improves documentation of a run but does not guarantee bitwise-identical GPU results across environments.
If a job fails during relaxation, the form recommends trying Relax Structure disabled. Inspect the inputs and any returned complexes rather than assuming that bypassing relaxation resolves a chemical or structural preparation problem.
Read rank, cLDDT, and predicted binding strength
Rank orders the sampled poses using the workflow's pose-quality assessment and clash considerations. Rank 1 is the preferred proposal among those samples, not proof of the experimental pose.
cLDDT is a dimensionless contact-confidence score on a 0–1 scale. Higher values indicate greater predicted support for local receptor–ligand contact geometry. It addresses structural plausibility rather than binding strength.
Binding Affinity is a learned affinity-style score where higher values indicate stronger predicted binding. It is not an energy in kcal/mol and should not be interpreted as a directly measured KD or IC50. Its direction differs from GNINA's empirical docking energy and Boltz-2's raw affinity value; do not combine their numbers into one ranking without a justified comparison method.
Inspect the affinity-versus-cLDDT plot for candidates supported by both geometry and the learned score. An attractive affinity-style value with weak pose confidence deserves investigation rather than automatic selection.
Compare receptor changes as well as ligand poses
Align the input and predicted proteins on a stable shared region. Examine pocket side chains, loop displacement, backbone distortions, and whether the proposed change conflicts with a neighboring domain or oligomer partner. Inspect ligand strain, buried unsatisfied polar groups, and clashes independently of the ranking.
Agreement across several samples can support a focused hypothesis. Multiple incompatible pocket arrangements indicate uncertainty that should be retained in the interpretation, especially when only one arrangement gives the desired interaction.
Example: investigate binding to an apo enzyme
Prepare the apo construct and a chemically verified ligand. Generate the default sample set, compare the top poses, and inspect which receptor residues move to accommodate binding. Identify contacts supported across plausible samples, then compare with available holo structures, mutagenesis evidence, or binding data.
If a candidate motivates GROMACS molecular dynamics, prepare a complete, parameterizable system first. A short stable MD trajectory can investigate local behavior, but it does not validate the docking pose or measure affinity by itself.
DynamicBind questions and troubleshooting
Does a predicted pocket opening prove a cryptic site exists? No. It proposes a conformation to investigate with independent structural and experimental evidence.
Why do ligands receive different receptor conformations? Ligand-specific rearrangement is part of the modeling task. Check whether changes are plausible and consistently supported, rather than assuming all movement is meaningful induced fit.
Can a high affinity score compensate for a distorted receptor? No. Review the receptor, ligand geometry, and contact confidence together before selecting candidates.
What is Neurosnap?
Neurosnap is the leading platform for bioinformatics and computational science focused on expanding access to powerful modeling and simulation tools. Because many state-of-the-art machine learning systems remain complex to install, configure, and scale, Neurosnap offers a clean, browser-based workspace that removes the burden of infrastructure management, dependency conflicts, and command-line tooling.
Built for biologists, chemists, and cross-disciplinary scientists, the platform enables advanced computational workflows without requiring expertise in software engineering or cloud architecture. Researchers can launch analyses through an intuitive interface, connect programmatically through a comprehensive API, and rely on automated resource management to scale workloads efficiently. By taking care of the underlying compute and operational complexity, Neurosnap allows teams to devote their energy to scientific progress and faster iteration. Security and data protection remain foundational principles, with clear safeguards outlined in our Terms of Use and Privacy Policy to ensure your work stays protected.
Advancing Discovery with DynamicBind on Neurosnap
Using DynamicBind on Neurosnap could drastically accelerate induced-fit protein-ligand docking from apo receptor structures with affinity-aware pose ranking.
- Apo-ready inputs: DynamicBind starts from an unbound receptor structure and a ligand, which matches many real discovery projects where no holo complex is available.
- Method-level advantage: The diffusion model can move both ligand and receptor, making it better suited to induced fit and cryptic-pocket problems than rigid docking alone.
- Practical controls:
Number Samples,Inference Steps,Model Version, andRelax Structurelet researchers choose between faster screening and broader conformational exploration. - Decision-ready ranking: Candidate complexes can be triaged by contact quality, predicted affinity, and cross-sample consistency before downstream simulation or wet-lab follow-up.
How to Use DynamicBind on Neurosnap
To harness the capabilities of DynamicBind, researchers can follow this streamlined workflow within Neurosnap:
- Access Neurosnap: Start by logging in to the Neurosnap website.
- Select Tool: From the list of available tools, choose DynamicBind.
- Provide Inputs: Provide all the inputs specified within the submission panel and optionally configure the tool as desired.
- Run Tool: Submit the DynamicBind job and Neurosnap will execute it in the cloud, automatically notifying you as soon as your results are ready.
- Review Output: Explore your results through rich visualizations, including figures, plots, and interactive views designed to help you analyze findings with clarity and confidence.
Citations
Please cite the original work when using DynamicBind in publications or research outputs.
|
Wei, L. et al. "DynamicBind: predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model", https://nature.com, 05 February 2024, https://www.nature.com/articles/s41467-024-45461-2. |
|
Neurosnap Inc. (2022). Neurosnap: An online platform for computational biology and chemistry. Available at: https://neurosnap.ai/ |
Similar Services
Explore related tools that support similar research workflows:
Proudly supporting 50,000+ scientists worldwide, including 7,000+ leading biotech and global biopharma organizations.
Making Scientific Research
Faster & Easier
Register for free — upgrade anytime.
Interested in getting a license? Contact Sales.
Try Free