Quick binding energy calculation with .bloom* platform

MM-GBSA combines molecular mechanics force fields with implicit solvation to rescore a docked complex. This guide walks through scoring your first complex, from importing a DiffDock result to interpreting the energies.

⚠︎ This implementation returns no binding free energy. It performs a single-point OpenMM evaluation of the minimized complex, so delta_g_bind is null, and there is no binding classification, no estimated Kd and no per-residue decomposition. Read how MM-GBSA results compare to the bench before acting on a result.

Step-by-step rescoring workflow

Access the MM-GBSA tool

Navigate to MM-GBSA in the .bloom* tool selector, or ask Ivy directly:

"Rescore my DiffDock poses with MM-GBSA"
"Calculate the binding energy of this complex"

Prepare your input complex

Option A: From DiffDock results (recommended)

Select a completed DiffDock job from the canvas or job panel
.bloom* automatically extracts the protein structure and docked ligand poses
Choose which poses to rescore: top 1 (fastest), top 5, or all poses

Option B: Manual upload

Upload a protein-ligand complex as a single PDB file containing both receptor and ligand
Or upload separate files: protein PDB and ligand SDF
.bloom* validates the structure and identifies the ligand automatically

Option C: From previous jobs

Select a docked complex from any previous docking job
Chain from a docking run (for example DiffDock) or any other step that produces protein-ligand complexes

Configure the solvent model

The force field is fixed at AMBER14 — there's no selector for it. What you can change:

Basic settings (recommended for first run):

Solvent model: gb_obc2 (default, best accuracy-speed balance)
Keep all other parameters at defaults

Advanced settings (via gear icon):

Solvent model: gb_obc2, gb_obc1, or gb_hct
Ionic strength: 0–1.0 mol/L (default: 0.15)
Minimize complex: Run energy minimization before evaluation (default: enabled)
Per-residue decomposition: Enable for binding hotspot identification (default: enabled)

Submit and monitor the job

Click Calculate Binding Energy to start the job
.bloom* shows real-time progress through parameterization, minimization, and energy evaluation phases
Typical completion: 1-5 minutes per pose on GPU
Multiple poses are scored sequentially

Review and analyze results

Energy summary:

Complex potential energy and nonbonded energy in kcal/mol, for the minimized complex
Generalized Born solvation term in kcal/mol
A caveat string stating what these are not. Read it
No DeltaG_bind, no binding classification and no estimated Kd — none of the three is computed. See the Kd question

Energy decomposition chart:

Van der Waals contribution (DeltaG_vdw)
Electrostatic contribution (DeltaG_elec)
Polar solvation penalty (DeltaG_polar)
Non-polar solvation contribution (DeltaG_nonpolar)
Stacked bar chart showing the relative magnitude of each term

Per-residue hotspot analysis:

Table of residues ranked by their contribution to binding energy
Negative values indicate binding-favorable residues (hotspots)
Positive values indicate residues that destabilize the complex
Residues highlighted on the 3D structure in the canvas viewer

Download options:

CSV: Full energy decomposition and per-residue data
JSON: Structured results with metadata
PDB: Minimized complex structure

Common rescoring scenarios

Single pose evaluation

Import the top DiffDock pose
Run MM-GBSA with the gb_obc2 solvent-model default
Check DeltaG_bind and binding classification
Review per-residue hotspots for mechanistic insight

Multi-pose ranking

Import all DiffDock poses (typically 5-10)
Run MM-GBSA on all poses
Sort by DeltaG_bind to identify the best-scoring pose
Compare energy decomposition across poses to understand binding mode differences

Compound series comparison

Dock a series of analogs with DiffDock
Run MM-GBSA on each compound's best pose
Rank compounds by DeltaG_bind
Use per-residue decomposition to understand SAR trends

Binding hotspot identification

Score a complex with per-residue decomposition enabled
Identify the top 5-10 contributing residues
Map hotspot residues onto the 3D structure
Use insights to guide mutagenesis experiments or analog design

Quality control checklist

Before submission

Docked complex has reasonable geometry (no steric clashes)
Ligand is in a reasonable binding pose (not solvent-exposed)
Protein has no missing residues in the binding site
Protonation states are appropriate (especially His, Cys, Asp, Glu)

After calculation

Read the caveat string before quoting any number from the result
Minimization converged (convergence_achieved)
Energy magnitudes are not wildly large — that signals a parameterization artifact, not tight binding
No parameterization warnings in the job log
The upstream pose has been inspected. A physics-based energy on a wrong pose is a precise answer to the wrong question

Next steps after rescoring

Compound triage:

Feed MM-GBSA energies into Compound Passport for comprehensive assessment
Compare energies with ADMET predictions to prioritize compounds
Ranking a compound series is what the method is for — but only within one target, and only once it emits a real ΔG_bind, which this implementation does not. Confirm affinity with ITC, SPR or an enzymatic IC50/Ki; see go / no-go

Mechanistic analysis:

Use per-residue hotspots to design mutations that test binding hypotheses
Analyze energy decomposition to understand whether binding is driven by shape complementarity (vdW) or electrostatics
Compare on-target vs. off-target binding energies for selectivity assessment

Further refinement:

Run molecular dynamics simulations starting from the MM-GBSA-minimized complex
For relative ranking within a congeneric series, fep_estimate gives a heuristic ΔΔG from descriptor deltas. It is not an alchemical free-energy calculation, so treat it as triage rather than a more rigorous alternative to MM-GBSA.
Use MM-GBSA hotspot data to guide scaffold decoration in GenMol

MM-GBSA quickstart