🔬 BLAST Identity Comparison Tool

This project compares the identity performance of BLASTn and BLASTx algorithms across DNA sequences of varying lengths and complexity.

---

📁 Project Structure

BLAST_ANALYSIS/
├── source_seq/             # Input .fasta sequences
│   ├── 1sequence.fasta
│   ├── 2sequence.fasta
│   └── ...
├── output_seq/             # All BLAST XML results (auto-generated)
│   ├── blastn_1sequence_100bp.xml
│   └── ...
├── pics/                   # Visualization output directory
│   ├── plot_algorithm_identity_avg.png
│   ├── plot_grouped_bar_length_algo.png
│   ├── plot_identity_compare.png
│   └── plot_length_vs_identity.png
├── venv/                   # Python virtual environment
├── blast_summary.csv       # Final summary table (auto-generated)
├── batch_run_blast_multi.py     # Main BLAST script
├── summarize_results.py         # Parses XML to generate CSV
├── parse_blast_results.py       # Optional: view top hits in terminal
├── plot_results.py              # Generates comparison graphs
├── BLAST_Colab.ipynb            # Google Colab notebook version

---

💻 Setup Instructions

1. Create and activate virtual environment (Windows)

python -m venv venv
.\venv\Scripts\activate

2. Install required libraries

pip install biopython pandas seaborn matplotlib

---

🚀 Run the Workflow

Step 1: Perform BLASTn and BLASTx queries

python batch_run_blast_multi.py

This reads source_seq/*.fasta, runs BLASTn/BLASTx at 100/500/1000bp, and saves results to output_seq/.

---

Step 2: Summarize results into CSV

python summarize_results.py

This parses all .xml results in output_seq/ and outputs blast_summary.csv.

---

Step 3: Generate visualizations

python plot_results.py

Creates graph files saved in the /pics directory:

• plot_length_vs_identity.png
• plot_algorithm_identity_avg.png
• plot_grouped_bar_length_algo.png

For the charts, we generate them locally via log files as we have a total of 2 devices, which reduces the production time. The images for this part are stored in the /pics directory.

---

📊 Sample Results Overview

• BLASTn achieves consistently high identity across all sequence lengths.
• BLASTx shows more variability, especially at 500bp.
• Longer sequences do not always improve match quality for BLASTx.

---

💻 Google Colab Implementation

Because the data is too large, the free Colab doesn't support us to train many times. We call the T100 graphics card to analyze and train in batches when the number of people using the Colab is small, and we also multitask in the same Colab notebook to fully utilize its performance.

The code provided here closely reflects our process, with only minor modifications to ensure reproducible results. Due to our multitasking approach in Colab, some execution paths may vary slightly.

Here is the Colab Notebook, which is also provided in this repository as BLAST_Colab.ipynb.

---

🧬 Data Sources

The sequences used in this analysis were obtained from the NCBI GenBank database:

• Human mitochondrion genome (NC_012920)
• E. coli genome (NC_000913.3)
• Mouse mitochondrion genome (NC_005089.1)
• Human BRCA1 gene (NG_005905.2)

Due to the large size of these files, they are not included directly in the repository.

---

📌 Conclusion

This tool helps explore when BLASTn or BLASTx performs better depending on sequence length and complexity. It can be easily extended to more sequences or to include BLASTp.

---

✍️ Author

CS123A Term Project
Spring 2025

Ray Zhang: rui.zhang@sjsu.edu
Jisheng Jiang: jisheng.jiang@sjsu.edu