Writing in STAT, Stanford's Euan Ashley describes asking Claude to analyze his whole genome using a framework his team developed in 2009. The AI completed in about 30 minutes and roughly $5 what once took nearly a year and 30 specialists. It re-identified major findings, including two copies of APOE e4, and flagged variants in DPYD and CYP2C19. It declined to calculate polygenic risk scores because the file lacked sufficient data.
Ashley stresses that a genome is not a fixed record. Older sequencing and reference-genome limitations can make interpretation misleading, especially in repetitive or structurally complex regions and in under-represented populations. He argues that as consumers gain access to AI genome analysis, the scientific community needs clear standards for what constitutes a high-quality medical interpretation.
Because this is a single opinion essay, it presents one author's experience and argument rather than peer-reviewed findings or independent verification.