Computational Modeling (NEW AREA - not yet covered by existing experiments)
Multiscale computational models integrating molecular dynamics, agent-based modeling, and machine learning can predict protein aggregation nucleation rates, elongation kinetics, and inter-cellular propagation patterns for tau, alpha-synuclein, and TDP-43, thereby identifying optimal intervention points for therapeutic development.
Current understanding of protein aggregation in neurodegenerative diseases is limited by:
- Inability to observe nucleation events in vivo
- Limited temporal resolution in human studies
- Species differences in animal models
- Lack of integration across scales (molecular → cellular → network → organism)
Computational modeling offers:
- Prediction of aggregation kinetics at timescales impossible to observe experimentally
- Virtual screening of mutation effects on aggregation propensity
- Integration of multi-omic data for personalized risk prediction
- Cost-effective hypothesis generation before experimental validation
Objective: Characterize monomer conformational dynamics and nucleation barriers
Approach:
- All-atom MD simulations of tau (PHF6 domain), alpha-synuclein (NAC region), and TDP-43 C-terminal domain
- Enhanced sampling (metadynamics, replica exchange) to capture rare conformational transitions
- Free energy calculations for dimerization and oligomerization
Reagents/Resources:
- Compute cluster: 5M GPU-hours (NVIDIA A100)
- Software licenses: GROMACS (open source), Desmond (Schrödinger - academic license)
- Personnel: 1 FTE computational biologist, 0.5 FTE biophysicist
Objective: Scale molecular kinetics to cellular and network levels
Approach:
- Agent-based model with protein aggregation states (monomer → oligomer → fibril → plaque)
- Incorporate cellular uptake, lysosomal trafficking, and exosome secretion
- Network model of spread between connected brain regions (structural connectome)
Reagents/Resources:
- Compute: 2M CPU-hours
- Software: Repast HPC (open source), custom Python framework
- Data: Allen Human Brain Atlas structural connectivity matrices
Objective: Predict individual patient trajectories and identify therapeutic targets
Approach:
- Train graph neural networks on multimodal patient data (genomics, proteomics, imaging)
- Integrate with Phase 1-2 models for personalized aggregation risk prediction
- Virtual knockouts to identify critical nodes in aggregation network
Reagents/Resources:
- Compute: 3M GPU-hours for training
- Data: UK Biobank, ADNI, PPMI datasets (access fees ~0K)
- Personnel: 1 FTE ML engineer
| Category |
Cost (USD) |
| Personnel (3 FTE × 12 months × $120K) |
$360,000 |
| Compute (GPU + CPU) |
$180,000 |
| Data access (UK Biobank, ADNI, PPMI) |
$50,000 |
| Software licenses |
$15,000 |
| Cloud storage/transfer |
$10,000 |
| Conference travel (2 domestic, 1 international) |
$8,000 |
| Publication fees (2 open-access) |
$6,000 |
| Total |
$629,000 |
- Month 1-2: Setup, data collection, baseline MD simulations
- Month 3-4: Enhanced sampling, dimerization free energies
- Month 5-6: Oligomerization kinetics, model parameterization
- Month 7-8: Agent-based model development
- Month 9-10: Network spread simulations, validation against human imaging data
- Month 11-12: ML integration, clinical prediction model, manuscript preparation
- Michele Vendruscolo (University of Cambridge) - Protein aggregation kinetics, computational approaches
- Tony Wyss-Coray (Stanford) - Aging, computational biology, data integration
- Marcus B. Jones (University of Chicago) - Tau propagation, computational models
- Viktor K. Sharma (NIH/NIA) - Alpha-synuclein computational modeling
- J. Alex B. McKenzie (UCL) - TDP-43 aggregation mechanisms
| Dimension |
Score (1-10) |
Rationale |
| Scientific Value (SV) |
9 |
Addresses fundamental question of how proteins aggregate and spread |
| Feasibility (F) |
8 |
Computational approach feasible with current infrastructure |
| Novelty (N) |
10 |
First integrated multiscale model for three proteins |
| Disease Impact (DI) |
9 |
Potential to identify novel therapeutic targets |
| Reach (R) |
8 |
Applicable across AD, PD, ALS, FTD |
| Cost Efficiency (CE) |
8 |
$630K for 3-year project is cost-effective |
| Time Efficiency (TE) |
7 |
12-month timeline is ambitious but achievable |
| Evidence Base (EB) |
7 |
Building on established MD and ABM methods |
| Addresses Uncertainty (AU) |
9 |
Directly addresses unknown nucleation mechanisms |
| Translation Potential (TP) |
8 |
Can inform clinical trial design and patient stratification |
Total Score: 83/140
- Validate Phase 1 predictions with in vitro ThT assays
- Calibrate agent-based model with longitudinal PET imaging data
- Apply ML model to predict clinical trial outcomes
This page provides information about Multiscale Computational Modeling of Protein Aggregation Kinetics.