Specialty & fine chemicals
DNA-personalised skincare
What candidate-gene skin panels actually measure, why common variants have small effect sizes, the one loss-of-function variant that genuinely is actionable, and the gap between an association and an ingredient recommendation.
A cheek swab, a marker panel, a report, a regimen. The logic is sound in outline, so the useful analysis is where each step’s strength actually sits.
What the panels read
Commercial skin panels are candidate-gene panels: a small, pre-selected set of common variants in genes with a plausible connection to skin biology. The recurring choices are consistent across providers.
Matrix turnover — MMP1 encodes a collagenase that degrades type I collagen; COL1A1 encodes the collagen itself. Oxidative defence — SOD2, CAT, GPX1 encode superoxide dismutase, catalase and glutathione peroxidase, the enzymes that dispose of reactive oxygen species. Pigmentation and UV response — MC1R variants shift melanogenesis toward pheomelanin, which is a weaker shield and itself photoreactive. Barrier — FLG encodes filaggrin, which aggregates keratin filaments and is then broken down into the amino acid derivatives that make up natural moisturising factor.
Every one of those is real biology with a real literature. That is the strength of the category, and it is where most of the marketing stops.
The problem is effect size, not existence
The variants typed are common polymorphisms, and common variants of complex traits almost always have small effects. A single-nucleotide change in a promoter or a coding region typically shifts a population risk by a small factor, not a categorical one.
Three consequences follow, and they are structural rather than fixable by a better panel.
Small effects do not survive individual prediction. An association measurable across thousands of people does not tell you where one person sits, because environment — cumulative UV dose, smoking, sleep, cleansing habit — accounts for far more of the observed variance in skin ageing than the typed panel does. Twin studies make this point directly: monozygotic twins with different sun exposure histories diverge visibly.
Panels of ten to forty markers cannot capture a polygenic trait. Where genome-wide studies have been run on skin phenotypes, they implicate many loci of individually tiny effect. A small candidate panel is a biased sample of that architecture, chosen for narrative plausibility.
The gene-to-ingredient step is the weakest link and is rarely evidenced. Suppose a SOD2 variant genuinely reduces enzyme activity. It does not follow that a topical antioxidant compensates: that requires the ingredient to penetrate, to reach the mitochondrion where SOD2 works, and to substitute functionally for an enzyme. Claims of tens of thousands of “gene–ingredient interactions” describe a database, not a body of intervention trials.
The exception worth naming
Filaggrin loss-of-function variants are different in kind. They are not common small-effect polymorphisms but null alleles that abolish protein production; they cause ichthyosis vulgaris and are the strongest known genetic risk factor for atopic dermatitis. Knowing a person carries one changes management in a way the rest of the panel does not — toward barrier repair, emollient discipline and irritant avoidance.
That is what an actionable result looks like, and it sets the standard against which the rest should be read.
The honest test
For a personalised recommendation to be worth its price, it must differ from the general one. Broad-spectrum sunscreen, a retinoid, barrier lipids and not smoking are the evidence-backed interventions for essentially everyone. Where a report arrives at those, the genotyping added confidence, not information — and that distinction is what a buyer is entitled to have stated.