A DIY biohacker sequenced his own genome five times with a benchtop nanopore rig

Bradley Woolf documents running his own DNA through an Oxford Nanopore MinION at home, taking a full end-to-end high-quality run from cheek swab to analysis five times over. The setup is not casual: the sequencer alone runs $7,500, plus a centrifuge, heat block, vortex, and a GPU-equipped machine for Dorado basecalling, on top of extraction kits, ONT ligation and flow-cell reagents, and Qubit quantification. He notes it took roughly two months to assemble everything, and that while costs remain out of reach for most people, they are falling exponentially. Cheek cells are the sample of choice because they are easy to collect and replenish quickly, though he’s clear about the limits — buccal swabs can’t tell you about cancer, localized inflammation, or tissue-specific gene expression, which require sampling the affected cells directly.

The more interesting argument is about what a personal genome is actually good for. Woolf frames the raw sequence as a static reference layer that only becomes useful once turned into a queryable VCF and run through annotation tools like VEP, ClinVar, gnomAD, and PharmGKB — or handed to an LLM such as Claude — to ask which variants you carry, which drug-metabolism pathways they affect, and where the models simply have no data yet. He’s careful to flag that this output is not diagnosis-grade and explicitly not a license to ‘edit yourself with CRISPR because an AI said so,’ while still betting that DNA (the stable reference) plus RNA (current state) plus biosensor data will eventually converge into a single real-time ‘model of yourself.’

The bulk of the piece is a dense, reproducible protocol: hardware and consumables lists, the full software stack (MinKNOW, Dorado, minimap2, samtools, Clair3, VEP), and step-by-step wet-lab instructions covering cell collection, pelleting, Monarch HMW lysis, bead-based DNA capture, and washing. Notably, he suggests feeding the protocol itself to an AI — even via AR glasses — to walk a newcomer through the bench work. It’s a snapshot of consumer genomics converging with LLM tooling, and a preview of the privacy and safety questions that follow when anyone can generate and query their own genetic data.