01
Overview
Nucleus Embryo allows parents to understand their future child’s health, personality, and physical characteristics based on what matters most to them. Every polygenic model behind Nucleus Embryo has been validated both across the population and within families in the UK Biobank, and the accuracy of each one is published on the Nucleus Labs page (1).
We offer the following analyses:
Women’s health: Endometriosis, polycystic ovarian syndrome (PCOS)
Nutrition and metabolic conditions: Alcohol dependence, celiac disease, type 1 diabetes, type 2 diabetes
Heart health: Coronary artery disease, hypertension
Cancers: Breast cancer, male breast cancer, ovarian cancer, prostate cancer, colorectal cancer
Neurological and mental health: ADHD, Alzheimer’s disease, anxiety disorders, autism spectrum disorder, bipolar disorder, depression, insomnia, migraine, multiple sclerosis, OCD, Parkinson’s disease, schizophrenia, IQ
Appearance and traits: Eye color, height, hair color, male-pattern baldness
Body and physical health: BMI, osteoarthritis, rheumatoid arthritis
Other conditions: Asthma, age-related macular degeneration, restless legs syndrome, seasonal allergies, severe acne
Hereditary disorders: With the Nucleus Preview test, you will learn if your future children could be at risk for over 2000 hereditary diseases like cystic fibrosis and Tay Sachs disease. If a risk is identified, a genetic counselor can discuss whether preimplantation genetic testing for monogenic disorders (PGT-M) is appropriate.
The calculator generates the relative risk reduction for diseases — and the difference in trait values — across two to ten viable embryos of European ancestry for the polygenic scores Nucleus Embryo offers (1). The figures shown are the expected difference between the lowest- and highest-scoring embryo in the set.
In 2025, we generated the most predictive published models for nine different diseases — Alzheimer’s disease, breast cancer, coronary artery disease, endometriosis, hypertension, prostate cancer, rheumatoid arthritis, type 1 diabetes, and type 2 diabetes — using the UK Biobank, Finngen Biobank, Million Veterans Program Biobank, and All of Us Biobank (2). These models, known as Nucleus Origin, were built from the genetic data of roughly 1.5 million people. They explain up to 22.9% of the variation in disease liability, matching or exceeding every previously published benchmark for each of the nine diseases. Three models- Alzheimer’s disease, prostate cancer, and type 2 diabetes- capture more than 75% of the risk that common genetic variants can explain (2). Validated in 40,872 siblings across 18,840 families, eight of the nine scores predicted disease as accurately within families as across the population (2).This set of open-weighted prediction models are publicly available for download.
We’ve since applied the same methodology to the rest of our disease models, and in 2026 we released Nucleus Vitruvian, our models for height, BMI, and intelligence (1, 3). In total, 35 polygenic models for diseases and traits are featured on Nucleus labs- each validated both across the population and within families in the UK Biobank (1). The data from these models are used in the above calculator to determine how much risk reduction can be expected for typical sets of embryos for parents of European ancestry. For example, if parents have 5 embryos, the embryo with the lowest risk of type 1 diabetes is expected to carry 83.8% less risk than the embryo with the highest risk, and 67% less risk than the average embryo in the set (2).
02
How It Works
With Nucleus Embryo, parents get their whole-genome sequenced with Nucleus Preview while clinics or prospective parents provide DNA data for embryos.
The biological parents will use Nucleus Preview to have their DNA analyzed for over 2000 hereditary diseases their children could be at risk for. By combining their Nucleus Preview results with their Nucleus Embryo analyses, users can ensure their embryos won’t have any rare, pathogenic markers that we analyze for. We then analyze the embryo DNA to determine their risks for the common diseases and traits listed above.
We only analyze the DNA of embryos with typical chromosome counts based on a preimplantation test known as PGT-A. That means users will receive analyses only on embryos that have already been screened for conditions like Down syndrome, Edward syndrome, and other conditions linked to chromosome number that can affect pregnancy success. We also analyze the DNA of embryos with mosaic or segmental chromosome changes identified through PGT-A.
03
Classical Mendelian analysis
Nucleus first analyzes the DNA from the biological parents for rare pathogenic and likely pathogenic variants — known as high-effect variants in our reports — in over 1800 genes to determine if they have any that could increase their future child’s risk for a disease. These include all 84 genes the American College of Medical Genetics and Genomics recommends reporting as secondary findings in clinical genome sequencing (4). You can find the growing list of genes we analyze here.
If any high-effect variants are identified, a genetic counselor can discuss whether preimplantation genetic testing for monogenic disorders (PGT-M) is appropriate to select for embryos lacking the variant(s). There may be cases, for example, where the identified variant(s) lead to a less severe disease. In these cases it may not make sense to include PGT-M.
Our team uses a combination of tools to identify consequential genetic variants, including the Ensembl Variant Effect Predictor, public and private databases, and guidelines from the American College of Medical Genetics, or the ACMG (4).
Our clinical geneticist ensures accurate, high-quality results by individually verifying each suspected pathogenic or likely pathogenic marker based on standards from the ACMG and the American Board of Bioanalysis.
Pathogenic variants are more likely than common genetic variants to have critical and immediate effects on the health of an individual patient or the future health of an embryo. Pathogenic markers supersede polygenic scores for any given condition in Nucleus reports.
For example, certain pathogenic markers in the LDLR gene have a disproportionately high impact on someone’s risk for heart disease (5). As a result, a polygenic score is less meaningful for assessing risk in such patients.
In these cases, Nucleus reports only the pathogenic marker — the highest-impact predictor of risk — to a patient or prospective parents.
04
Polygenic risk score analysis
In the absence of a pathogenic or likely pathogenic variant, we provide polygenic scores, commonly known as PGS, for adults and embryos. We offer these scores, which we refer to as common genetic scores, for the common diseases and traits noted above.
Common genetic scores are calculated by considering the combined impact an embryo’s common genetic variants have on their future disease risk or trait expression. The effect of each common variant is measured from large genome-wide association studies, or GWAS (5). GWAS are large population studies that uncover how common DNA differences, usually single nucleotide polymorphisms, or SNPs, are associated with a particular trait or disease.
The exact source of the GWAS association data differs for each report. The most predictive models are the ones that Nucleus generates using a wide range of databases including the UK Biobank, All of Us Research Program, FinnGen, the Million Veteran Program, and others (1, 3). This is explained in greater detail in Folkersen et al. (2020) (6), but the guiding principle is to always provide the most recent, well-established, and predictive scores for each disease and trait.
Nucleus calculates polygenic scores by first determining the number of risk alleles an embryo has that are associated with a particular disease or trait. This number is then multiplied by each respective variant’s impact on a phenotype, known as its effect size.
Polygenic models add these effect sizes together, resulting in a single number that measures someone’s genetic disposition for a phenotype based on the combined influence of the common genetic markers in a score.
Nucleus estimates these effect sizes with SBayesRC, a Bayesian method that jointly models over 7 million common variants together with 96 functional genomic annotations, rather than the roughly 1 million variants used by most published scores (3, 7). In benchmarking, SBayesRC improved prediction accuracy by 14% over its predecessor and outperformed other leading methods, with the largest gains in non-European ancestries (7).
Nucleus delivers common genetic scores on a bell curve with an average of 0 and a standard deviation of 1. The distance between someone’s risk and the average risk of a person with their genetic ancestry is known as a Z-score that generally ranges from -4 to 4.
The most predictive polygenic models will come from large, ancestrally diverse datasets containing millions — or even billions — of genomes. Today, however, most GWASs are calculated with data from European populations.
To help account for this gap, Nucleus first standardizes genetic scores across ancestries. This is done based on allele frequencies from the 1000 Genomes Project (7, 8). We next modulate the polygenic score’s estimated predictive power according to the genetic distance of the user’s ancestry from the ancestry of the training population to ensure non-European populations can still receive rigorous, meaningful genetic analysis results today (9). Nucleus also validated Origin models directly in the multi-ancestry All of Us cohort, including 52,023 people of African, 49,519 of Admixed-American, 6,410 of East Asian, and 3,111 of South Asian ancestry (2).
As such, common genetic scores from Nucleus indicate the relative increase in genetic risk compared to people of the same ancestry, and the absolute risk numbers take decreased predictivity in non-European populations into account.
05
Applying PGS to embryonic prediction
There is strong evidence that polygenic scores can reflect true genetic influences, not artifacts from environment or population stratification.
Research shows that polygenic scores for diseases are less likely to be impacted by factors that could confound predictions, like assortative mating, where people tend to marry those with similar characteristics (10). Furthermore, research shows that the heritability of these phenotypes doesn’t change across relatives and people who are unrelated, indicating true direct genetic effects (11). Recent research also shows, compared to behavioral genetics phenotypes like IQ, clinical phenotypes like migraines have negligible indirect effects (12).
Indeed, studies have shown that polygenic scores can meaningfully distinguish disease risk even among siblings raised in the same environment — supporting their ability to capture real genetic influence.
For example, in sibling pairs where one had a much higher polygenic score than the other, the higher-risk sibling was the one affected by diseases like breast cancer, diabetes, or heart disease in up to 90% of cases (13, 14).
The same was true for trait predictions like height, a strong model that correctly identified the taller sibling in up to 80% of cases.
Nucleus applies the same sibling test to its own scores. In the Nucleus Origin study, the within-family predictive power of eight of the nine disease scores was statistically indistinguishable from their population-level power. The one exception, type 1 diabetes, showed modest attenuation and its predictive power was reduced accordingly (2). Nucleus reports the same within-family validation for every model it offers, alongside its population-level accuracy, on the Nucleus Labs page (1).
One study found that when choosing between two and five viable embryos, selecting embryos with lowest polygenic scores for a complex disease can result in substantial reduction of an embryo’s relative risk for the disease (15). Using the Nucleus Origin scores, a couple with five viable embryos can expect a 27% to 67% reduction in relative risk across the nine diseases by selecting the lowest-risk embryo rather than an average one — 55% for Alzheimer’s disease, 54% for prostate cancer, 50% for type 2 diabetes, and 67% for type 1 diabetes (1).
The same is true for traits like height and intelligence. A study of couples of European ancestry showed that, with the scores available in 2019, polygenic scores for height could capture up to a 2.5 centimeter difference across five embryos, while scores for intelligence could capture 2.5 IQ pointswere expected to add about 2.5 centimeters, and scores for intelligence about 2.5 IQ points, by selecting the highest-scoring of five embryos over the average embryo (16). This translates to around 5 centimeters and 5 IQ points when comparing the highest- and lowest-scoring of five embryos.
Building on this research, Nucleus Labs’ (Nucleus’ research arm) height and IQ models can lead to average gains of around 3.1 inches for height and around 10.8 IQ points for intelligence when selecting from five embryos.Nucleus Vitruvian — the height, BMI, and intelligence models developed by Nucleus Labs, Nucleus’ research arm — explains 43% of the variation in height, 18% of the variation in BMI, and 19% of the variation in intelligence within families (3). The expected difference between the highest- and lowest-scoring of five embryos is around 3 inches (7.5 centimeters) in height, around 10.8 IQ points, and around 5.2 kg/m² in BMI (3). That is roughly 1.5 times the height difference and twice the IQ difference projected in 2019, and larger than what the best academic scores available today can capture, even though those scores were trained on up to nearly four times as many people (3).
Across the body of its models, Nucleus’ scores match or exceed the most predictive previously published polygenic scores for the same diseases and traits, whether developed by academic consortia or by industry (1, 3). Nucleus is also the only provider to release its disease models as open weights, and it publishes the population and within-family accuracy of every score it offers so that these comparisons can be checked independently (1).
06
Responsible embryonic polygenic prediction
The nature of most genetic prediction is one of uncertainty. For common diseases and traits, genetics alone will never be able to predict an outcome. Put simply, DNA is not destiny.
Nucleus takes great care in delivering complex, probabilistic results to parents, while respecting their right to access a wide range of genetic analyses.
The limit of genetic influence on any outcome is known as broad-sense heritability, or H². Broad-sense heritability is a percentage that indicates how much of the variation in any given disease or trait can theoretically be explained by DNA (17). Polygenic models today predict a fraction of the heritability of any given phenotype. This fraction — a number that encapsulates the degree of genetic influence that can be captured today — is known as the model’s variability explained.
Every embryo analysis Nucleus offers is clearly labeled a strong, modest, or weak predictor of a genetic outcome based on the model’s variability explained, contextualizing the genetic predictions they receive from Nucleus to inform their selection strategy. In addition, a set of analyses are available that are at the research stage and still have limited predictiveness.
Nucleus further contextualizes polygenic scores by reflecting their variability explained in the absolute scores we provide. The higher a genetic model’s variability explained, the more accurately, on average, Nucleus can predict a phenotype in an embryo.
For example, genetics can explain 80% of the variation in height among the population. Nucleus considers its height model a strong predictor, able to predict 42.8% of height variance within families, over about half of that variation — 43% of height variance, measured within families (15321). As a result, Nucleus can predict an embryo’s expected height with more accuracy than any other continuous trait, like IQ or BMI.
Alternatively, the research into the genetics of ovarian cancer is still at an early stage. While scientists estimate the broad-sense heritability of ovarian cancer to be 39%, today’s models capture around 2.6% of that risk (1, 18, 20).
Nucleus also strives to generate and provide users with the most highly predictive models available today — which includes those that are least likely to conflate non-genetic influences for heritable ones.
That said, developing polygenic scores that isolate genetic influences that directly affect an outcome from non-genetic ones is an active area of research. Researchers have found that models for ADHD and IQ are among the most likely to conflate genetic and non-genetic influences, resulting in less applicable predictions in the context of embryonic selection (12). Nucleus validates each model within-family to most accurately capture the direct genetic effects, and reports the resulting within-family accuracy next to the population accuracy for every model (20). For example, Nucleus’ intelligence model explains 23.8% of variation within the population and 19.1% within families, and the within-family figure is used for embryo predictions (1, 3).
Our offerings will continue to evolve as our understanding of embryonic polygenic prediction improves.
07
Why we made Nucleus Embryo
We provide transparent and clear science in a digestible format that powers informed choices for individual health and that of future generations.
Nucleus Embryo generates accurate and understandable information in its detailed reports by indicating which diseases and traits can be most accurately predicted at this time, and by providing graphs, charts, and references for each condition or trait.
These kinds of insights are in high demand, and more parents are embracing new technology to have healthy children who thrive. A wide-ranging study of Americans found the majority accepted the use of genetic technology to choose embryos based on health and personality traits (19). Four in 10 parents would use genetic optimization as another tool to increase their child’s chances of going to a top college.
One in 50 people in the U.S. are conceived with IVF. We believe that the advanced analyses Nucleus Embryo can provide will help parents deeply understand their choices when it comes to their future child — sparking a new level of choice in reproductive care, just like IVF first did for hopeful parents decades ago.
References
1
Nucleus Genomics. “Nucleus Labs: Genetic Optimization Hub — UK Biobank Validation of Nucleus Models.” mynucleus.com/labs. Accessed September 2026.
2
Cordogan S, Starr DB, Treff N, et al. (2025) Within- and Between-Family Validation of Nine Polygenic Risk Scores Developed in 1.5 Million Individuals: Implications for IVF, Embryo Selection, and Reduction in Lifetime Disease Risk medRxiv. doi: https://doi.org/10.1101/2025.10.24.25338613 (current version v3, February 2026).
3
Nucleus Genomics Scientific Team. Polygenic prediction of height, BMI, and intelligence within families: Revisiting the limited utility of embryo screening. Nucleus Labs white paper, 2026. mynucleus.com/labs/traits-whitepaper.
4
Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015 May;17(5):405-24. doi: 10.1038/gim.2015.30. Epub 2015 Mar 5. PMID: 25741868.
5
Ison HE, Clarke SL, Knowles JW. Familial Hypercholesterolemia. 2014 Jan 2 [Updated 2022 Jul 7]. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993-2024.
6
Lee SH, van der Werf JH, Hayes BJ, et al. Predicting unobserved phenotypes for complex traits from whole-genome SNP data. PLoS Genet. 2008 Oct;4(10):e1000231. doi: 10.1371/journal.pgen.1000231. Epub 2008 Oct 24. PMID: 18949033.
7
Folkersen L, Pain O, Ingason A, et al. Impute.me: An open-source, non-profit tool for using data from direct-to-consumer genetic testing to calculate and interpret polygenic risk scores. Front Genet. 2020 Jun 30;11:578. doi: 10.3389/fgene.2020.00578. PMID: 32714365.
8
Zheng Z, Liu S, Sidorenko J, et al. Leveraging functional genomic annotations and genome coverage to improve polygenic prediction of complex traits within and between ancestries. Nat Genet. 2024 May;56(5):767-777. doi: 10.1038/s41588-024-01704-y. PMID: 38689000.
9
1000 Genomes Project Consortium; Auton A, Brooks LD, Durbin RM, et al. A global reference for human genetic variation. Nature. 2015 Oct 1;526(7571):68-74. doi: 10.1038/nature15393. PMID: 26432245.
10
“The International Genome Sample Resource” InternationalGenome.org.
11
Privé F, Aschard H, Carmi S, et al. Portability of 245 polygenic scores when derived from the UK Biobank and applied to 9 ancestry groups from the same cohort. Am J Hum Genet. 2022 Jan 6;109(1):12-23. doi: 10.1016/j.ajhg.2021.11.008. Erratum in: Am J Hum Genet. 2022 Feb 3;109(2):373. PMID: 34995502.
12
Horwitz TB, Balbona JV, Paulich KN, et al. Evidence of correlations between human partners based on systematic reviews and meta-analyses of 22 traits and UK Biobank analysis of 133 traits. Nat Hum Behav. 2023 Sep;7(9):1568-1583. doi: 10.1038/s41562-023-01672-z. Epub 2023 Aug 31. PMID: 37653148.
13
Howe LJ, Nivard MG, Morris TT, et al. Within-sibship genome-wide association analyses decrease bias in estimates of direct genetic effects. Nat Genet. 2022 May;54(5):581-592. doi: 10.1038/s41588-022-01062-7. Epub 2022 May 9. PMID: 35534559.
14
Tan T, Jayashankar H, Guan J, et al. Family-GWAS reveals effects of environment and mating on genetic associations. medRxiv. doi: https://doi.org/10.1101/2024.10.01.24314703.
15
Lello L, Raben TG, Hsu SDH. Sibling validation of polygenic risk scores and complex trait prediction. Sci Rep. 2020 Aug 6;10(1):13190. doi: 10.1038/s41598-020-69927-7. PMID: 32764582.
16
Widen E, Lello L, Raben TG, et al. Polygenic Health Index, General Health, and Pleiotropy: Sibling Analysis and Disease Risk Reduction. Sci Rep. 2022 Oct 28;12(1):18173. doi: 10.1038/s41598-022-22637-8. PMID: 36307513.
17
Lencz T, Backenroth D, Granot-Hershkovitz E, et al. Utility of polygenic embryo screening for disease depends on the selection strategy. Elife. 2021 Oct 12;10:e64716. doi: 10.7554/eLife.64716. PMID: 34635206.
18
Karavani E, Zuk O, Zeevi D, et al. Screening Human Embryos for Polygenic Traits Has Limited Utility. Cell. 2019 Nov 27;179(6):1424-1435.e8. doi: 10.1016/j.cell.2019.10.033. Epub 2019 Nov 21. PMID: 31761530.
19
“Estimating Trait Heritability” Scitable.
20
Fritsche LG, Patil S, Beesley LJ, et al. Cancer PRSweb: An Online Repository with Polygenic Risk Scores for Major Cancer Traits and Their Evaluation in Two Independent Biobanks. Am J Hum Genet. 2020 Nov 5;107(5):815-836. doi: 10.1016/j.ajhg.2020.08.025. Epub 2020 Sep 28. PMID: 32991828.
21
Meyer MN, Tan T, Benjamin DJ, et al. Public views on polygenic screening of embryos. Science. 2023 Feb 10;379(6632):541-543. doi: 10.1126/science.ade1083. Epub 2023 Feb 9. PMID: 36758092.
Keep exploring
Codex session expired. Sign in again.