# Understanding AlphaFold Output Files

This document explains the output files generated by AlphaFold and their significance for protein structure prediction.

## Output Directory Structure

The output directory contains several types of files for each model prediction:

### 1. Model Predictions

- `unrelaxed_model_[1-5]_pred_0.pdb`: Initial predicted structure in PDB format

- `unrelaxed_model_[1-5]_pred_0.cif`: Initial predicted structure in mmCIF format

- `relaxed_model_[1-5]_pred_0.pdb`: Relaxed (energy-minimized) structure in PDB format

- `relaxed_model_[1-5]_pred_0.cif`: Relaxed structure in mmCIF format

### 2. Confidence Metrics

- `confidence_model_[1-5]_pred_0.json`: Per-residue confidence scores

  - Contains three key arrays:

    - `residueNumber`: Position in the sequence

    - `confidenceScore`: pLDDT score (0-100)

    - `confidenceCategory`: Confidence category (L=Low, M=Medium, H=High)

  - pLDDT interpretation:

    - < 50: Low confidence (L)

    - 50-70: Medium confidence (M)

    - 70-90: High confidence (H)

    - > 90: Very high confidence (H)

### 3. Ranking and Analysis

- `ranked_[0-4].pdb/.cif`: Models sorted by confidence, with 0 being the best

- `ranking_debug.json`: Details about the model ranking process

- `relax_metrics.json`: Metrics from the structure relaxation step

- `timings.json`: Performance metrics for different stages of prediction

### 4. Intermediate Data

- `features.pkl`: Input features extracted from the sequence

- `result_model_[1-5]_pred_0.pkl`: Raw prediction results

- `msas/`: Directory containing multiple sequence alignments used for prediction

## Key Files for Analysis

1. **Best Model**: Always check `ranked_0.pdb` first - this is AlphaFold's best prediction

2. **Confidence Assessment**: Review `confidence_model_[1-5]_pred_0.json` to understand prediction reliability

   - High confidence scores (>90) suggest very reliable predictions

   - Lower scores may indicate flexible or disordered regions

## File Formats

1. **PDB Files** (.pdb):

   - Standard format for protein structures

   - Easily viewable in molecular visualization software (PyMOL, VMD, etc.)

   - Contains atomic coordinates and basic metadata

2. **mmCIF Files** (.cif):

   - Modern format for protein structures

   - More detailed than PDB format

   - Better handles large structures and contains more metadata

## Using the Output

1. **Structure Analysis**:

   - Use `ranked_0.pdb` for your primary analysis

   - Compare with other ranked models to assess structural variability

   - Pay attention to regions with high confidence scores

2. **Quality Assessment**:

   - Check confidence scores to identify reliable regions

   - Look for consistently high-confidence regions across models

   - Be cautious about interpreting low-confidence regions

3. **Visualization**:

   - Color structure by pLDDT score to highlight reliable regions

   - Compare multiple models to understand structural flexibility

   - Focus analysis on high-confidence regions

## Visualizing PDB Files

To visualize the predicted protein structures (PDB files), you have several options:

1. **Desktop Applications**:

   - **PyMOL**: Professional-grade molecular visualization (recommended)

     - Download from: https://pymol.org/

     - Features:

       - High-quality rendering

       - Powerful analysis tools

       - Script automation support

       - Can color by B-factor (confidence scores in AlphaFold)

   - **UCSF Chimera**: Free academic visualization tool

     - Download from: https://www.cgl.ucsf.edu/chimera/

     - Good for structure analysis and comparison

   - **VMD**: Specialized in molecular dynamics

     - Download from: https://www.ks.uiuc.edu/Research/vmd/

     - Excellent for trajectory analysis

2. **Web-Based Viewers**:

   - **Mol***: Modern web-based viewer

     - Access via: https://molstar.org/viewer/

     - Just drag and drop your PDB file

   - **NGL Viewer**: Lightweight web viewer

     - Access via: http://nglviewer.org/ngl/

     - Good for quick visualization

3. **Analyzing ranked_0.pdb**:

   - This is AlphaFold's best prediction

   - The B-factor column (last number in each ATOM record) contains pLDDT confidence scores

   - In PyMOL, you can color by B-factor to visualize confidence:

     ```python

     # PyMOL commands

     load ranked_0.pdb

     spectrum b, rainbow   # Colors structure by confidence scores

     ```

4. **Best Practices**:

   - Always start by viewing `ranked_0.pdb`

   - Color the structure by confidence scores

   - Compare multiple models to understand flexibility

   - Look for regions with high confidence scores (>90)

   - Be cautious about interpreting low-confidence regions

5. **Important Features to Look For**:

   - Secondary structure elements (helices, sheets)

   - Overall fold and domain organization

   - Regions of high vs. low confidence

   - Potential flexible regions (varying between models)

   - Biologically important sites or motifs

## Understanding Pickle (.pkl) Files

AlphaFold generates two types of pickle files that contain detailed prediction data:

1. **result_model_[1-5]_pred_0.pkl**:

   Contains the raw prediction outputs including:

   - `distogram`: Distance predictions between residue pairs

   - `experimentally_resolved`: Predictions about atom positions

   - `masked_msa`: Multiple sequence alignment information

   - `predicted_lddt`: Raw predictions for local confidence

   - `structure_module`: Final atom positions and masks

   - `plddt`: Per-residue confidence scores (0-100)

   - `ranking_confidence`: Overall model confidence

2. **features.pkl**:

   Contains input features used for prediction:

   - `sequence`: The input protein sequence

   - `aatype`: Amino acid type encodings

   - `msa`: Multiple sequence alignment data

   - `template_*`: Information about structural templates

   - `residue_index`: Numbering of residues

   - `domain_name`: Name of the protein domain

### Key Metrics from Pickle Files

1. **Confidence Scores (pLDDT)**:

   - Range: 0-100

   - Higher is better

   - Your results show:

     - Average score: 89.16 (Very good)

     - Range: 65.29 - 93.71

     - Most residues have high confidence (>70)

2. **Multiple Sequence Alignment (MSA)**:

   - Your protein had 3,303 aligned sequences

   - This is a good number for prediction accuracy

   - More diverse alignments generally improve prediction quality

3. **Structure Module**:

   - Contains final atomic coordinates

   - Includes all backbone and side chain atoms

   - Shape: (34 residues, 37 atoms per residue, 3 coordinates)

### Analyzing Pickle Files

To analyze these files, you can use the provided `inspect_pkl.py` script:

```python

python inspect_pkl.py

```

This will show:

- Data structure of predictions

- Confidence scores

- Sequence information

- Template usage

- MSA statistics

## Why Five Models?

AlphaFold generates five models for each prediction for several important reasons:

1. **Sampling Different Conformations**:

   - Proteins can exist in multiple stable conformations

   - Different models may capture different possible structural states

   - Helps identify flexible or dynamic regions of the protein

2. **Confidence Assessment**:

   - Agreement between models indicates prediction reliability

   - Regions that vary between models may be:

     - Naturally flexible

     - Have multiple possible conformations

     - Harder to predict accurately

3. **Model Architecture**:

   - Each model uses slightly different neural network parameters

   - Models are trained independently with different random seeds

   - This ensemble approach improves prediction robustness

4. **Ranking System**:

   - AlphaFold ranks the five models based on predicted confidence

   - `ranked_0` represents the most confident prediction

   - Comparing ranks helps identify the most likely structure

5. **Scientific Best Practice**:

   - Multiple models follow the scientific principle of ensemble sampling

   - Helps avoid over-relying on a single prediction

   - Provides error estimates for the prediction

When analyzing results, it's important to:

- Start with the highest-ranked model (`ranked_0`)

- Compare models to identify consistent and variable regions

- Consider all models when the confidence scores are similar

## Performance Metrics

The `timings.json` file provides detailed information about:

- MSA generation time

- Feature processing time

- Model prediction time

- Structure relaxation time

This can be useful for:

- Optimizing future runs

- Understanding computational requirements

- Identifying bottlenecks in the prediction pipeline