PGS-TRI estimates genetic effects and environmental interactions using family data on autism and orofacial clefts
Background: Why distinguish genetic risk within families
A statistical method called PGS-TRI has been introduced to estimate the direct effects of polygenic scores, differences between parental indirect effects, and interactions with the environment using genetic data from affected children and both parents. Applied to family data on autism spectrum disorder (ASD) and orofacial clefts (OFC), it showed different patterns of association depending on the condition and trait. The study was published in Nature Genetics on June 2, 2026.
A polygenic score (PGS) combines information from multiple genetic variants to indicate genetic predisposition to a particular trait. Comparisons among unrelated people can be affected by population structure or assortative mating, in which people with similar characteristics choose one another as partners. Within-family comparisons are therefore used as an alternative. Researchers also examine whether parental genetic characteristics may be associated with a child’s health through pathways other than the variants transmitted to the child.
The existing polygenic transmission disequilibrium test (pTDT) assesses whether an affected child’s PGS deviates from the parental average, but it has limitations in providing effect sizes on an appropriate risk scale. PGS-TRI estimates relative risks by adjusting this difference for within-family PGS variance. It also evaluates gene–environment interactions by appropriately adjusting the covariance between environmental factors and the transmission disequilibrium statistic.
Key findings: Effects and differences observed in two developmental conditions
The researchers analyzed 18,383 ASD child–parent trios from SPARK and 1,904 OFC trios from GENEVA. Simulations accounting for population structure and assortative mating generally showed estimates with little bias and appropriate false-positive rates and confidence intervals. However, when mean PGS differed by parental sex, additional adjustment was needed to estimate differences in indirect effects.
Among 12,813 European-ancestry ASD trios, the relative risk for the direct effect per one-standard-deviation increase in ASD PGS was 1.28 (95% confidence interval, 1.25–1.31). In the full family dataset, the strength of the association with ASD PGS declined as genetic distance increased from the European-ancestry population used to train the score. This illustrates a limitation in applying scores across populations. The difference itself, however, does not imply a biological mechanism.
For the ASD PGS itself, the researchers found no evidence of a difference between maternal and paternal indirect effects. Differences were observed for parental PGSs for body mass index (BMI) and some neurocognitive traits. The quantity estimated was the difference between the two parental indirect effects, rather than each parent’s indirect effect separately. Selection bias arising from differences in study participation by parental sex could explain some of the signals.
In the OFC analysis, existing PGSs were directly associated with the risk of cleft lip without cleft palate and cleft lip with cleft palate. No evidence of association was found for cleft palate alone, consistent with the fact that the scores were developed for subtypes that include cleft lip. Limited evidence of interactions, including with maternal smoking during pregnancy, was observed in some European-ancestry subtypes.
In ASD, there was no strong evidence of interaction with the prenatal and perinatal factors examined. This means there was insufficient evidence that genetic and environmental effects departed substantially from a multiplicative pattern on the relative-risk scale; it does not mean that environmental factors have no effect.
Implications and outlook: Uses and interpretive limits of risk estimation
PGS-TRI is a tool for estimating the magnitude of genetic associations in family data and for checking findings from existing population studies. Observed associations cannot be interpreted directly as preventive effects or treatment mechanisms. Further validation is needed before the method can be used in parental counseling or targeted preventive interventions.
Interactions between environmental factors and parental indirect effects were not included in this model, and the model assumptions about the relationship between genetic predisposition and environmental exposures must also be considered. Another open question is whether gene-expression and metabolite scores trained using adult data adequately reflect molecular traits in childhood. The paper made its analysis software and code available, while access to individual participant data requires a separate process. The study design and results can be found in the original paper: https://www.nature.com/articles/s41588-026-02601-2
Nature Genetics, Published June 2026.
Summary: Bypassing the analytical blind spots of conventional genome-wide association screening—which frequently miscalculates human disease risk by decoupling individual direct genetic variants from parental nurture background nodes—this study introduces a structural statistical framework utilizing parent-offspring trios. Configured to capture multi-layered inheritance matrices, the computing platform simultaneously estimates direct genetic effects, indirect parental genetic pathways, and non-linear gene-environment ($G \times E$) interaction kinetics within a single multidimensional tensor model. The unified platform eliminates population stratification anomalies, systematically clarifying hidden missing heritability parameters across diverse disease registries, and providing a scalable computational baseline for predictive multivariate polygenic risk score (PRS) calibration, tailored environmental containment, and clinical translational screening.
Researchers studying developmental conditions can analyze the PGSs of affected children and both parents alongside pregnancy exposure data to distinguish and assess the direct effect of inherited genetic predisposition, the difference between maternal and paternal indirect effects, and environmental interactions. In OFC research, the method may help test in follow-up cohorts whether the effects of existing scores vary by disease subtype or maternal smoking status. It can also be used to check whether score associations hold in families whose genetic backgrounds differ from those of the training population. The study did not test individual-level diagnostic accuracy or the effectiveness of preventive interventions. Signals involving parental indirect effects also do not establish causal effects of specific parenting behaviors or parental responsibility; selection bias and actual mediating pathways need further investigation before applying the findings in counseling.