Exposome researchers are still developing the standards needed to pair exposure data with genomes

That 70 to 90 per cent figure rests on a single reference cited in the feature ^[5]. The clinic visit the feature imagines for 2050 would combine blood, urine, a postal code and a description of habits with a genome to tally one person's disease risks ^[1]. For that job, a share of risk is the wrong kind of number. The visit needs an effect size measured in individuals, with a genome-only model as the control, and the feature does not report one.

The biology is better worked out. "Genetics load the gun, but the environment pulls the trigger," said Francis Collins, former director of the US National Institutes of Health ^[6]. The feature describes an epigenetic route: exposures can switch genes on or off, and tobacco smoke can silence tumour-suppressor genes ^[7].

Measurement is harder. "The signatures of past exposures stay in the body," said Gary Miller, director of the Center for Innovative Exposomics at Columbia University ^[10]. Those signatures include sugars, proteins and chemical breakdown products in blood or urine ^[16]. A durable mark like smoking's ^[11] can be read from one blood draw. Exposures that change over time, which the feature calls dynamic, need repeated reads, and researchers are still building the capacity to monitor them continuously ^[17].

The thing a faster mass spectrometer doesn't tell you is how to join its output to everything else. The feature's list of required sources takes in biobanks, long-term cohort studies, surveys of diet, medication and physical activity, and geospatial models of air pollution ^[9]. The shared standards and protocols the feature calls much-needed are still being developed ^[15].

Attention is growing faster than that infrastructure. This year's PubMed count for "exposome" is already about 3.2 times the total for all of 2020 ^[1], and the year is not finished ^[14].

Of the two objections from sceptics that the feature reports ^[13], I think the claim that the totality of human exposure cannot be measured matters more for medicine. The feature's own working standard concedes part of it, asking a rigorous study to assess as many non-genetic disease factors as possible ^[9]. Miller's answer is about care. "If one does not address lifestyle factors, diet, medications, environmental exposures and other factors, care is not personalized," he said ^[8]. Christopher Wild, who coined the term in 2005 ^[3], wrote of a "desperate need to develop methods with the same precision for an individual's environmental exposure as we have for the individual's genome" ^[4].