News · 2026-09-28

Nucleus's Vitruvian score reports embryo trait ranges, not a guaranteed IQ gain

Nucleus's Vitruvian release reports polygenic embryo scores for height, body-mass index and an intelligence measure, but its widely repeated 14.3 IQ-point figure is a high-to-low range across ten hypothetical embryos—not a promised increase in a chosen child's IQ. That distinction is central because professional and advertising bodies say the clinical and consumer claims remain unproven or inadequately substantiated.

Key facts

• Release: Nucleus says Vitruvian trait scores were introduced in a 2 September 2026 update.
• Method: the whitepaper reports within-family analysis of roughly 40,000 siblings from about 20,000 UK Biobank sibling pairs.
• Anchor statistic: Nucleus reports 19% variance explained for its intelligence measure and a 14.3-point high-to-low spread for ten hypothetical embryos.
• Caveat: ASRM says trait selection should not be used; NAD challenged relevant advertising claims.

Nucleus says it uses SBayesRC with about seven million imputed common variants and 96 functional annotations, drawing on FinnGen, the Million Veteran Program and UK Biobank. Its intelligence outcome is the first in-person UK Biobank verbal-numerical reasoning test, adjusted for test-retest reliability. The company reports variance explained of 43% for height, 18% for BMI and 19% for the intelligence measure in its within-family analysis.

The statistical translation is the story. For a hypothetical five-embryo cohort, the paper gives an expected highest-to-lowest spread of about 10.8 IQ-scale points. It says the top-score versus cohort-average advantage is half that expected spread. For ten embryos, the high-to-low spread rises to 14.3. Saying “choose an embryo and gain 14 IQ points” swaps a range between hypothetical extremes for an expected outcome of a selected individual. It is like saying that the height gap between the tallest and shortest contestant in a group is the height boost received by choosing one contestant.

That does not make the underlying work imaginary. Nucleus has a named company, a technical document and a released consumer product concept. It does mean the whitepaper's model output, the uncertainty in embryo ranking and a prediction about an eventual child are three different steps. The company itself provides a consumer deep dive that uses “around 10.8 IQ points” in a way that can blur those steps; the technical paper is the better anchor.

The strongest counterweight comes from clinical and advertising institutions. ASRM's ethics opinion calls PGT-P nascent and unproven, says it should remain research-only pending evidence, and says selecting for height or intelligence is outside reproductive medicine and should not be used. It flags ancestry bias, gene-environment complexity, limited embryo counts and unestablished clinical utility. NAD's September decision recommended that Nucleus modify or discontinue claims implying it could identify the smartest embryo or validate the claimed magnitude of differences; Nucleus appealed and voluntarily discontinued one broader potential claim.

The caveat is not “statistics are useless.” Within-family methods are an attempt to make a difficult prediction problem more realistic than population-only correlations. But a score can be calibrated in one dataset and still fail the clinical standard for offering consequential selection decisions to families. This is a sharp example of calibration and confidence: a model can produce a number with a technical interpretation without the number being a warrant for a life-changing decision.

The durable lesson is precision in language. Vitruvian's release is a real commercial and scientific development in polygenic prediction. The evidence does not support presenting a score spread as a guaranteed IQ enhancement, and the relevant medical and advertising authorities explicitly dispute the leap from model output to consumer claim.

Primary source, verified: read the paper →

Ground Truth. (2026, September 28). Nucleus's Vitruvian score reports embryo trait ranges, not a guaranteed IQ gain. Ground Truth. https://groundtruth.day/news/nucleus-vitruvian-embryo-iq-score-claim.html

@misc{groundtruth:nucleus-vitruvian-embryo-iq-score-claim,
  title  = {Nucleus's Vitruvian score reports embryo trait ranges, not a guaranteed IQ gain},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {sep},
  url    = {https://groundtruth.day/news/nucleus-vitruvian-embryo-iq-score-claim.html}
}

Topics: health-ai · genomics · policy · evaluation · ethics