Version 0.0.1
A purely computational, physics-based framework for predicting protein corona formation on nanoparticle surfaces using DEDIK (Discrete Event Dynamics with Implicit Kinetics).
DEDIK (Discrete Event Dynamics with Implicit Kinetics) is a simulation method that captures protein corona dynamics without explicit rate constants.
Instead of solving differential equations with k_on/k_off:
Traditional Mean Field:
dθ/dt = k_on × C × (1-θ) - k_off × θ
↑ ↑
Need experiments for these
SurfCo DEDIK:
For each iteration:
1. Select protein (weighted by concentration C)
2. Select event type (adsorption/desorption)
3. If adsorption: Check collisions with present proteins occupying the NP surface
4. Calculate acceptance probability from binding energies E
5. Accept/reject based on probabilities
6. Update system state
- Concentration → Selection Frequency: Higher concentration = more binding attempts (mimics arrival/collision rate)
- Energy → Acceptance Probability: Stronger binding = higher acceptance (mimics k_on/k_off ratio)
- Result: Mimics realistic dynamics without explicit rate constants!
Mimics temporal evolution - From empty surface to equilibrium
Kinetic competition - Multiple proteins competing for surface
Vroman effect - Sparse strong binders displace abundant weak binders over iteration progression
Thermodynamic equilibration - System minimizes energy
Realistic dynamics - Comparable to mean field predictions and experimental results
- High geometric accuracy (0.1 nm voxel resolution)
- Full 3D volumetric simulation
- Memory-efficient (sparse tracking)
- Database caching (avoid recomputation)
- Parallel processing during structure generation
- Interactive 3D visualizations
- Operating System: Linux/Unix
- Python: 3.8 or higher
- Memory: 8 GB minimum
- CPU: Multi-core processor (optional)
- Disk Space: 1-5 GB for typical projects
┌─────────────────────────────────────────────┐
│ INPUT: Protein structures (PDB) + │
│ Nanoparticle properties + │
│ Concentrations + DEDIK config │
└──────────────────┬──────────────────────────┘
│
▼
┌─────────────────────────────────────────────┐
│ MODULE 1: Energy Calculator │
│ • Run UnitedAtom for each protein │
│ • Calculate binding energies │
│ • Find optimal orientations │
│ • Normalize to probability range │
└──────────────────┬──────────────────────────┘
│
▼
┌─────────────────────────────────────────────┐
│ MODULE 2: 3D Structure Generator │
│ • Coarse-grain proteins (amino acids) │
│ • Position on nanoparticle surface │
│ • Pre-compute 72 rotational variants │
│ • Calculate spatial extents │
└──────────────────┬──────────────────────────┘
│
▼
┌─────────────────────────────────────────────┐
│ MODULE 3: DEDIK Simulation │
│ • Initialize 3D voxel grid │
│ • Iterative loop: │
│ - Select protein (by concentration) │
│ - Propose random event │
│ - Check spatial collisions │
│ - Accept/reject by energy │
│ - Allow competitive displacement │
│ • Evolve to equilibrium │
└──────────────────┬──────────────────────────┘
│
▼
┌─────────────────────────────────────────────┐
│ MODULE 4: Visualization │
│ • 3D interactive corona structure │
│ • Time series plots │
│ • Composition analysis │
└─────────────────────────────────────────────┘
| Aspect | Mean Field Models | SurfCo DEDIK |
|---|---|---|
| Rate constants | Required (k_on, k_off) | Not required |
| Experiments | Case specific experiments needed | None needed |
| Parameter source | Experimental fit | Computed from structures |
| Algorithm | Differential equations | Discrete event simulation |
| Spatial detail | Mean field (averaged) | Full 3D explicit geometry |
| Dynamics | Temporal evolution | Mimics temporal evolution |
| Vroman effect | Yes (with multi-component) | Yes (emergent from physics) |
| Scalability | Limited by experiments | Purely computational |
| Novel materials | Need new experiments | Direct computation |
| Throughput | ~1 system/month (lab) | ~10-100 systems/day (compute) |
Key Point: Both show realistic protein corona dynamics, but DEDIK achieves this through pure computation rather than experimental parameterization.
Energy-Based Probabilities:
- Strong binders (E < 0): High p_adsorb, Low p_desorb → Accumulate
- Weak binders (E > 0): Low p_adsorb, High p_desorb → Transient
Concentration Weighting:
- Abundant proteins selected more often
- Mimics higher collision frequency with surface
- No explicit rate constant needed!
Spatial Constraints:
- 3D voxel grid prevents overlaps
- Penetration correction ensures physical validity
- 0.1 nm resolution for high accuracy
Competitive Dynamics:
- Strong binders can displace weak binders
- Energy difference determines displacement probability
- Vroman effect emerges naturally
SurfCo simulates the temporal evolution of protein corona formation through discrete event dynamics. The framework captures realistic protein competition and adsorption/desorption kinetics and showcases results similiar to current computational models (Mean-Field models) and experimental results.
The system in the example of SurfCo (Figure 1) is using the same proteins, the same protein concentrations (in this variation we used the two monomers of Fibrinogen to maintain their realistic structures so we cut their concentration by half) and the closest NP material to the system used by " Dell'Orco, Daniele, et al. "Modeling the time evolution of the nanoparticle-protein corona in a body fluid." PloS one 5.6 (2010): e10949. " (Figure 2).
one can see that both the Mean-Field model results and the experimental results presented in this paper match the phenomena and trends captured by SurfCo (Figures 1 and 2). more data and results of SurfCo for this system can be found in the projects directory inside "example" directory.
Figure 1: Protein corona formation and stabilization as iterations progress on a 35 nm radius carbon black nanoparticle, the proteins in the system are HSA, HDL, FIB-A and FIB-B. Using the DEDIK model.
Figure 2: Protein corona formation and stabilization as time progress on a 35 nm radius carbon black nanoparticle, the proteins in the system are HSA (red), HDL (blue), Fibrinogen (green). Using a Mean-Field model.
SurfCo is developed at the Dana Research Group (Technion).
SurfCo's contributors are:
- Benjamin Shugaev (leading developer)
- Dr. Alon Grinberg Dana