How dellacortelab/prospr works
Main Components
The system should be divided into these key components:
a) Input Processing Layer
• A3M file input (Multiple Sequence Alignment)
• Data loading and preprocessing components (from prospr/dataloader.py)
b) Neural Network Core
• Three neural network models (prospr0421_a.pt, prospr0421_b.pt, prospr0421_c.pt)
• Neural network architecture (defined in prospr/nn.py)
c) Prediction Pipeline
• Prediction workflow (prospr/prediction.py)
• Multiple output types (distance maps, ASA, phi/psi angles, secondary structure)
d) Training Infrastructure
• Training pipeline (prospr/train.py)
• Data flow from training datasets
Component Relationships
• Training data flow from data/training directory
• Evaluation data flow from data/evaluate directory
• Output flow to data/predictions and data/results
Specific Elements
┌─────────────────────────────────────┐
• Input Processing │
• A3M file handling │
• Sequence processing │
• ┘
• ↓
┌─────────────────────────────────────┐
• Neural Network Ensemble │
• Model A (prospr0421_a.pt) │
• Model B (prospr0421_b.pt) │
• Model C (prospr0421_c.pt) │
• ┘
• ↓
┌─────────────────────────────────────┐
• Prediction Output │
• Distance Maps │
• ASA Predictions │
• Phi/Psi Angles │
• Secondary Structure │
• ┘
Additional Details
Elements:
• Data persistence layers (training data, model weights, predictions)
• Evaluation pipeline
• Training/Prediction mode switch
Technical Specifications
Annotations for:
• File formats (.a3m, .pdb, .pkl)
• Model architecture details
• Input/output specifications
• Performance metrics
Training mode (for model development)
This will create a comprehensive system design that accurately represents ProSPr's architecture while maintaining clarity and usefulness for development and documentation purposes.