GenSLM-TrpB

KO | EN

gitlite — search

#javascript #ai #go #golang #android #nodejs #java #windows #awesome #kubernetes #security #pytorch #mcp #ios

(★ 10)

Sequence-based Generative AI Guided Design of Versatile Tryptophan Synthase

GenSLM-TrpB Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# GenSLM-TrpB

Sequence-based Generative AI Guided Design of Versatile Tryptophan Synthase

![Demo](https://raw.githubusercontent.com/AI-ProteinDesign/GenSLM-TrpB/main/data/230_TrpB.gif)

<sub>**Artificially generated enzyme: 230**</sub>

## Overview

This repository contains a comprehensive pipeline for filtering and analyzing protein sequences generated by GenSLM (Generative Sequence Language Model) for TrpB (Tryptophan Synthase Beta subunit) design. The pipeline includes multiple filtering steps to ensure quality and relevance of generated sequences.

## Repository Structure

```
GenSLM-TrpB/
├── filters/                    # Filtering pipeline scripts
│   ├── esm_plddt_filter.py    # ESM-2 pLDDT score filtering
│   ├── length_check.py        # Sequence length validation
│   ├── run_blast_pairwise.py  # BLAST pairwise alignment
│   ├── parse_blast_output.py  # Parse BLAST results
│   ├── run_global_alignment.py # Global sequence alignment
│   ├── check_blast_clusters.py # BLAST clustering analysis
│   └── ...                    # Additional utility scripts
├── model/                      # Model files and weights
│   └── genslm_25m/           # GenSLM 25M parameter model
├── selected_candidates.fasta  # Final filtered sequences
└── README.md                  # This file
```

## Prerequisites

- Python 3.7+
- Required Python packages (see `requirements.txt` for versions):
  - BioPython
  - NumPy
  - Pandas
  - SciPy
  - PyTorch
  - Transformers
  - Matplotlib
  - Seaborn
  - Jupyter (for visualization)
  - BLAST+ (for sequence alignment)
  - MAFFT (for multiple sequence alignment)

## Installation

1. Clone the repository:
```bash
git clone <repository-url>
cd GenSLM-TrpB
```

2. Install required Python packages:
```bash
# Install from requirements.txt (recommended)
pip install -r requirements.txt

# Or install packages individually
pip install biopython numpy pandas matplotlib jupyter torch transformers scipy seaborn tqdm scikit-learn
```

3. Install BLAST+ and MAFFT (if not already installed):
   - **Ubuntu/Debian**: `sudo apt-get install ncbi-blast+ mafft`
   - **macOS**: `brew install blast mafft`
   - **Windows**: Download from NCBI website for BLAST+, use conda for MAFFT: `conda install -c bioconda mafft`

## Data Generation

Before running the filtering pipeline, you need to generate protein sequences using the GenSLM model.

### Generate Sequences with GenSLM

```bash
cd model
python generate_genslm.py --total_seq 1000 --checkpoint checkpoints/sft_stability/checkpoint-9375 --output_fasta name_of_file.fasta
```

**Parameters:**
- `--total_seq`: Number of sequences to generate (e.g., 1000)
- `--checkpoint`: Path to the trained model checkpoint
- `--output_fasta`: Output FASTA file name for generated sequences

**Example:**
```bash
python generate_genslm.py --total_seq 1000 --checkpoint checkpoints/sft_stability/checkpoint-9375 --output_fasta generated_trpB_sequences.fasta
```

**Note:** Ensure you have the required model checkpoints in the `model/checkpoints/` directory before running generation.

## Filtering Pipeline

The filtering process follows a specific sequence to ensure quality and relevance of generated protein sequences. Follow these steps in order:

### Step 1: ESM-2 pLDDT Score Filtering
```bash
cd filters
python esm_plddt_filter.py
```
**Purpose**: Filters sequences based on ESM-2 predicted pLDDT scores to ensure structural quality.

### Step 2: Length Validation
```bash
python length_check.py
```
**Purpose**: Validates that sequences meet length requirements for TrpB proteins.

### Step 3: BLAST Pairwise Alignment
```bash
# Set working directory in run_blast_pairwise.py for storing BLAST results
python run_blast_pairwise.py [reference_proteins.fasta] [generated_proteins.fasta]
```
**Purpose**: Performs pairwise BLAST alignment between generated and reference sequences.

### Step 4: Parse BLAST Output
```bash
python parse_blast_output.py [output.pkl]
```
**Purpose**: Extracts maximum identity information from BLAST results.
**Output**: Pickle file containing tuples: `(gen_seq_id, ref_seq_id, maxID)`

### Step 5: Global Sequence Alignment
```bash
python run_global_alignment.py [blast_pickle.pkl] [generated_proteins.fasta] [reference_proteins.fasta] [alignment_output.pkl]
```
**Purpose**: Performs global sequence alignment for detailed similarity analysis.
**Output**: Pickle file containing tuples: `(gen_seq_id, ref_seq_id, maxID, alignment_score)`

### Step 6: Generate Filtering Statistics
```bash
jupyter notebook filtering_stats.ipynb
```
**Purpose**: Creates visualizations and statistical analysis of the filtering results.

### Step 7: BLAST Clustering Analysis
```bash
python check_blast_clusters.py
```
**Purpose**: Analyze BLAST results to identify sequence clusters and patterns.

## Input Files

- **Reference Proteins**: FASTA file containing known TrpB sequences for comparison
- **Generated Proteins**: FASTA file containing sequences from the GenSLM model
- **Working Directory**: Directory for storing intermediate BLAST results

## Output Files

- **Filtered Sequences**: High-quality sequences that pass all filters
- **BLAST Results**: Pairwise alignment scores and identity metrics
- **Alignment Scores**: Global alignment results with similarity metrics
- **Statistics**: Visualizations and analysis of filtering results
- **Clusters**: Sequence clustering analysis results

## Notes

- Ensure all input FASTA files are properly formatted
- Set appropriate working directories for BLAST results storage
- Monitor intermediate pickle files for data integrity
- The pipeline is designed to be run sequentially for optimal results

## Troubleshooting

- **BLAST errors**: Ensure BLAST+ is properly installed and in PATH
- **Memory issues**: Large sequence files may require increased memory allocation
- **File paths**: Use absolute paths or ensure relative paths are correct
- **Python packages**: Verify all required packages are installed and up-to-date

## Citation

If you use this pipeline in your research, please cite the relevant papers for GenSLM and the filtering methods used.

## License

This project is licensed under the MIT License – see the LICENSE file for details.

## Contact

Contact Amin Tavakoli (amint@caltech.edu) with any questions or inquiries.

GenSLM-TrpB

GenSLM-TrpB Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

GenSLM-TrpB

GenSLM-TrpB Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README