Wiem-rhouma/dna-unsupervised-analysis

Unsupervised genomic analysis of African populations using PCA and clustering # 🧬 Reading History from DNA Without Labels ## Overview This project explores whether population history can be recovered using only DNA data without labels. ## Dataset - 995 individuals - 7 populations (YRI, GWD, ESN, LWK, MSL, ACB, ASW) - ~5000 SNPs after preprocessing ## ⚙️ Pipeline 1. Data loading & cleaning 2. SNP encoding (0/1/2 dosage) 3. MAF filtering 4. PCA (dimensionality reduction) 5. Clustering (K-means) 6. Admixture detection 7. IBS similarity analysis 8. Sex-based analysis ## Key Results - PCA reveals population structure without labels - Clear diaspora vs continental separation - Admixture patterns detected in ACB & ASW populations - No bottleneck detected, diversity explained by mixing ## Tech Stack - Python - Pandas, NumPy - Scikit-learn - Matplotlib, Seaborn ## Files - `src/DNA.py` → main pipeline - `report/` → full report - `presentation/` → slides ## How to run ```bash pip install -r requirements.txt python src/DNA.py