biogen_analytics

Bioinformatics pipeline for genomic analysis of FASTA sequences: DNA/protein metrics calculation, NCBI homology search, statistical modeling with linear regression, and anomaly detection.

---

Description

This project analyzes human DNA sequences (FASTA format) by applying a 4-phase pipeline:

1. Genomic Processing — GC%, length, DNA→protein translation, molecular weight, aromaticity.
2. Online Homology (optional) — BLAST search against the NCBI nt database and reference downloading.
3. AI Model — Linear regression (length → GC%) with anomaly detection (>2σ).
4. Visualization — Correlation heatmap generated with seaborn.

The included sample genes are: TP53, BRCA1, APOE, and KCNJ1.

---

Project Structure

text
Biogen_analytics/
├── data/
│   ├── raw/                  # Input FASTAs (.fasta / .fa)
│   ├── processed/            # features_genes.csv (generated)
│   └── reference/            # References downloaded from NCBI (online mode)
├── graficos/                 # heatmap_correlacion.png (generated)
├── logs/                     # pipeline.log (generated)
├── resultados/               # Results CSV and PNG (generated)
├── src/
│   ├── __init__.py           # CLI entry point
│   ├── descarga_sec.py       # Bulk download of FASTAs from NCBI
│   └── fetch_tools.py        # Reference FASTA downloading from NCBI
│   ├── homology_search.py    # BLAST + GenBank details extraction
│   ├── model_ia.py           # Linear regression + anomaly detector
│   ├── procesado_sec.py      # DNA/Protein metrics orchestrator
│   └── visualizer.py         # Correlation heatmap with seaborn
├── tests/
│   └── test_pipeline.py      # 29 unit tests (pytest)
├── .env.example              # Environment variables template
├── requirements.txt
├── main.py                   # Orchestrator: DNA + protein metrics
└── setup.py                  # Folder initialization script

---

Installation

git clone [https://github.com/tu_usuario/BioGen-Predictive-Pipeline.git](https://github.com/tu_usuario/BioGen-Predictive-Pipeline.git)
cd BioGen-Predictive-Pipeline

python -m venv venv
# Windows:
venv\Scripts\activate
# Linux/Mac:
source venv/bin/activate

pip install -r requirements.txt

Configure NCBI Credentials

Copy .env.example to .env and add your email:

cp .env.example .env

ENTREZ_EMAIL=tu_email@ejemplo.com

An email address is required for NCBI API calls. Without it, the --online mode will fail.

---

Usage

Offline Mode (Local analysis, no NCBI)

python scripts/main_pipeline.py

Online Mode (BLAST + reference download)

python scripts/main_pipeline.py --online

---

Tests

pytest tests/ -v

Expected output: 29 passed.

---

Tech Stack

Library	Usage
BioPython ≥1.81	FASTA parsing, BLAST, Entrez, ProteinAnalysis
pandas ≥2.0	DataFrames and CSV export
scikit-learn ≥1.3	Linear regression
seaborn / matplotlib	Correlation heatmap
python-dotenv	Secure credentials management
pytest	Unit testing

---

Dataset

The sequences included in data/raw/ are public RefSeq records from NCBI:

Gene	Accession	Function
TP53	NM_000546	Supresor tumoral
BRCA1	NM_007294	Reparación de ADN
APOE	NM_000041	Metabolismo lipídico
KCNJ1	NM_001301717	Canal de potasio

---

Author

Carlos Garcia Corona