Amyloid-β 42 (Aβ42)
System Overview
Conformational state analysis of the Amyloid-β 42 monomer using GraphVAMPNets. Identifies metastable states, transition kinetics between conformational ensembles, and residue-level attention patterns.
Open Interactive Report
Opens the full-page interactive visualization with 3D embeddings, protein structure viewer, transition matrices, and attention patterns.
Pipeline Parameters
Preparation
Graph Construction
- k-Nearest Neighbors
- 4
- Nodes
- 42
- Edges
- 168
- Node Feature Dim
- 42
- Edge Feature Dim
- 16
- Node Embedding Dim
- 16
- Gaussian Expansion Dim
- 16
Graph2Vec Embedding
- Embedding Dim
- 512
- Epochs
- 50
- Max Degree
- 3
- Min Count
- 10
- UMAP Dims Tested
- 2, 3, 5, 6, 7, 10
State Discovery
- Recommended States
- 9
- Selection Method
- max_across_metrics
- Chosen Source
- umap_10
- Elbow k
- 3
- Silhouette Recommendation
- 2
- Elbow Recommendation
- 3
- Bic Recommendation
- 9
- Aic Recommendation
- 9
- Cluster Sizes
- 16.7%, 15.1%, 12.0%, 11.7%, 11.5%, 11.4%, 8.0%, 6.8%, 6.8%
- Min States Tested
- 2
- Max States Tested
- 10
Training
Network Architecture
- Encoder Type
- schnet
Embedding MLP
- Enabled
- True
- Hidden Dim
- 64
- Output Dim
- 32
- Layers
- 2
- Dropout
- 0.0
- Activation
- relu
Classifier Head
- Hidden Dim
- 64
- Layers
- 2
- Dropout
- 0.1
- Activation
- relu
- Normalization
- batch_norm
Hyperparameters
- Epochs
- 100
- Batch Size
- 32
- Learning Rate
- 0.001
- Weight Decay
- 0.0001
- Validation Split
- 0.2
Analysis
Trained Models
- lag10.0ns_9states
- best_model.pt
- lag5.0ns_9states
- best_model.pt
Completed Analyses
- lag10.0ns_9states
- completed
- lag5.0ns_9states
- completed
Run Info
- Timestamp
- 2026-02-12T15:15:48.750095