DrTim105/genes-and-geography

Replicating and extending Novembre et al. (2008) — end-to-end Python pipeline for human population structure analysis using 1000 Genomes Phase 1 chr22 data, with a within-Africa extension, Fst-based ancestry-informative marker identification, and an interactive Streamlit explorer. # Genes and Geography > **Recovering global population structure and identifying ancestry-informative markers from 1000 Genomes chromosome 22 data — a Python pipeline replicating and extending Novembre et al. (2008).** --- ## TL;DR Using ~5,000 SNPs from chromosome 22 across 1,092 individuals, this project: - **Recovers four continental population clusters** via PCA, with the African cluster ~43% more diverse than the European cluster — the genomic signature of the out-of-Africa bottleneck. - **Identifies the specific variants encoding that structure** using Fst, finding that just **50 variants (1% of the data) reproduce the same clustering** as the full panel. - **Connects the result to clinical medicine** — the same kind of high-Fst variants underlie why drug dosing algorithms calibrated in European populations fail in Nigerian patients. --- ## Why This Matters (For Non-Specialists) Every human carries roughly 3 billion DNA letters. At ~10 million positions, people differ from each other — sometimes one person has a "T" where another has a "C". These differences are called **variants**, and most of them are biologically silent. But the pattern of variants you carry tells a story about where your ancestors came from, because populations that lived apart accumulated slightly different patterns over thousands of years. This project takes DNA data from 1,092 people across Africa, Europe, East Asia, and the Americas, and asks a simple question: **if you map their genetic variation onto a 2D plot, does the plot look anything like a real-world map?** The answer is yes. People cluster by geography on the plot — Yoruba (Nigeria) on one side, Han Chinese (Beijing) on another, British (England) in between. Nobody told the algorithm where anyone is from. It discovered geography from the DNA alone. The project then goes further. It quantifies how much more genetic variety African populations carry compared to Europeans (43% more, on chromosome 22 alone). It id …