Predictive genotype-to-trait models for complex traits
Predicting a complex trait - height, cognition, disease risk, temperament - accurately enough from a genome to choose or design variants on purpose.
State of the art
Embryo-screening companies offer polygenic predictions for disease risk, height, BMI and IQ (2025-2026); within-family validation shows far smaller effects than population-level scores imply.
- Scores tag correlated variants rather than causal ones, so editing the top-ranked variant need not move the trait
- Non-additive effects, pleiotropy and gene-environment interaction are unmodeled
- Ancestry portability is poor because training cohorts are predominantly European
Who’s working on it
- Shai CarmiResearcherHebrew University of Jerusalem
Quantified the small expected gain from polygenic embryo selection for height and IQ (2019) and the selection-strategy dependence of disease-risk reduction (eLife, 2021)
- Genomic PredictionCompany
Introduced the first clinical polygenic embryo test in 2019 and still supplies the LifeView disease-risk product used by IVF clinics
- Orchid HealthStartup
Palo Alto company selling whole-genome embryo reports restricted to health risks rather than cognitive or cosmetic traits as of 2026
- HerasightStartup
Began selling embryo predictions for height, BMI, longevity and IQ in 2025, the furthest any vendor has gone beyond disease risk
- Nucleus GenomicsStartup
Offers embryo screening for traits including eye colour, hair colour and handedness; vendors collectively reported scoring thousands of embryos by 2026
- UK BiobankProgram
The half-million-participant cohort nearly every published polygenic score is trained on, and the reason those scores transfer poorly outside European ancestry
What to invest in
| Kind | Investment | Scale |
|---|---|---|
| research | Within-family and multi-ancestry biobank cohorts at million scale with functional follow-up to separate causal from tagging variants | $$$ $100M – $1B |
| research | Perturbation atlases linking edited variants to measured cellular and organismal phenotypes | $$ $10M – $100M |
Depends on
A root node: no prerequisites modeled.