TL;DR: On 22 October 2026, researchers Chao Ning and Xiang Zhou published fastGxE, a genome-wide method that finds genetic variants whose effects change with environment and runs 32.98–126.49 times faster than existing approaches.

What was published on 22 October 2026

A Technical Report titled “Genome-wide detection of genotype–environment interactions at biobank scale with fastGxE” appeared in Nature Genetics on 22 October 2026, according to Nature Genetics.

The report lists Chao Ning and Xiang Zhou as authors, according to Nature Genetics.

Both authors are affiliated with the Department of Biostatistics and the Center for Statistical Genetics at the University of Michigan in Ann Arbor, according to the preprint record on PMC.

Correspondence is directed to Xiang Zhou at [email protected], according to the preprint record on PMC.

The Nature Genetics article is open access, according to Nature Genetics.

The work was supported in part by the National Institutes of Health, according to the preprint record on PMC.

The preprint version was posted to Research Square on 20 March 2025 with DOI 10.21203/rs.3.rs-5952773/v1 under a Creative Commons Attribution 4.0 International License, according to the preprint record on PMC.

The authors declared no competing interest in that preprint record, according to the preprint record on PMC.

The published method is named fast genotype–environment interaction analysis, abbreviated fastGxE, according to Nature Genetics.

Traditional genome-wide association studies have primarily focused on detecting main genotype effects, according to Nature Genetics.

That traditional focus often overlooks genotype–environment interactions, abbreviated GxE, according to Nature Genetics.

GxE describes genetic variants that interact with environmental factors to influence traits of interest, according to Nature Genetics.

Understanding those interactions is presented as essential for understanding context-specific genetic effects and refining disease etiology, according to Nature Genetics.

Why context-specific effects were missed

Genome-wide association studies have identified many genetic variants associated with diseases and disease-related phenotypes, according to Nature Genetics.

Those discoveries reshaped understanding of the genetic architecture underlying complex traits, according to Nature Genetics.

Most of those studies focused on genetic main effects, according to Nature Genetics.

A main-effect analysis implicitly assumes a constant effect size across environmental contexts, according to Nature Genetics.

That assumption overlooks the possibility that genetic effects can be context-dependent and modulated by environmental factors, according to Nature Genetics.

Both genetic and environmental factors are major contributors to phenotypic variation, according to Nature Genetics.

Their interactions play a pivotal role in shaping complex human traits, according to Nature Genetics.

The paper illustrates the point with obesity, where genetic predisposition may manifest differently depending on lifestyle factors such as diet and physical activity, according to Nature Genetics.

Large-scale biobank data have provided an unprecedented opportunity to systematically characterize GxE at a scale and resolution previously unattainable, according to Nature Genetics.

Leveraging those datasets, GxE studies have started to dissect the dynamic interplay between genes and the environment, according to Nature Genetics.

The same studies aim to uncover context-specific genetic effects, according to Nature Genetics.

They also aim to generate deeper insights into the architecture of complex traits, according to Nature Genetics.

They aim to build more accurate models for genetic risk prediction by incorporating GxE effects, according to Nature Genetics.

They ultimately aim to pave the way for personalized medicine strategies that integrate both genetic and environmental factors, according to Nature Genetics.

Two older strategies and their limits

The literature describes two primary strategies for detecting GxE associations, according to Nature Genetics.

The first strategy focuses on single-environment GxE tests, according to Nature Genetics.

Tools for that strategy include the linear regression framework implemented in PLINK, according to Nature Genetics.

They also include the linear mixed model framework implemented in fastGWA-GE, according to Nature Genetics.

These single-environment approaches directly assess interaction between a genetic variant and a single environmental factor, according to the preprint record on PMC.

They often lack sufficient statistical power, even in large-scale biobank data, according to Nature Genetics.

They have sometimes failed to detect well-established GxE loci, according to Nature Genetics.

One example is the FTO locus for body mass index, abbreviated BMI, which a single-environment analysis failed to detect, according to Nature Genetics.

The alternative strategy involves joint GxE testing across multiple environmental factors, according to Nature Genetics.

Implementations include StructLMM and GEM, according to Nature Genetics.

These multi-environment approaches offer much improved statistical power compared with single-environment tests, according to Nature Genetics.

The advantage reflects the observation that genetic variants often interact with multiple environmental factors to influence phenotypes, according to Nature Genetics.

Despite that advantage, the multi-environment tools have three key limitations, according to Nature Genetics.

First, statistical power can be substantially reduced when only a small subset of environments contributes to GxE interactions, according to Nature Genetics.

Second, the tools are computationally intensive, according to Nature Genetics.

They have been restricted to analyzing candidate genomic regions, according to Nature Genetics.

They lack the scalability required for genome-wide GxE analysis in biobank-scale datasets, according to Nature Genetics.

Third, these approaches do not account for polygenic effects arising from genome-wide variants, according to Nature Genetics.

They also do not account for polygenic interaction effects arising from genome-wide variants, according to Nature Genetics.

They also do not model heterogeneous noise variance across environments, according to Nature Genetics.

Both types of polygenic effects explain a substantial proportion of phenotypic variance, according to Nature Genetics.

Failing to account for them can lead to excessive false positives in GxE association analyses, according to Nature Genetics.

Ignoring heterogeneous noise can bias estimates of the underlying genetic architecture, according to Nature Genetics.

How fastGxE is designed

fastGxE is described as a scalable and effective genome-wide multi-environment GxE method, according to Nature Genetics.

It is designed to identify genetic variants that interact with one or multiple environmental factors, according to Nature Genetics.

It examines one SNP at a time, according to the fastGxE Overview.

It decomposes the phenotype into SNP main effect, environmental main effects, and GxE interaction effects, according to the fastGxE Overview.

While doing so, it controls for polygenic effects, according to the fastGxE Overview.

It also controls for polygenic interaction effects, according to the fastGxE Overview.

It also controls for noise heterogeneity, according to the fastGxE Overview.

The Nature Genetics abstract lists the same three controls: polygenic effects, polygenic interaction effects, and noise heterogeneity, according to Nature Genetics.

By explicitly modeling polygenic background and heteroscedastic noise, fastGxE generates calibrated p-values for identifying candidate GxE loci, according to the fastGxE Overview.

The method evaluates various GxE effect-size configurations, according to the fastGxE Overview.

It combines the resulting p-values into a single p-value, according to the fastGxE Overview.

That combined p-value tests whether the SNP interacts with at least one environmental factor, according to the fastGxE Overview.

The design remains robust to the number of environmental factors involved in GxE interactions, according to Nature Genetics.

It also remains robust to the number of interacting environments, according to Nature Genetics.

The authors illustrate the benefits through extensive simulations, according to Nature Genetics.

They also illustrate the benefits through in-depth analysis of UK Biobank data, according to Nature Genetics.

mmSuSiE pinpoints driving environments

The team also developed mmSuSiE, according to Nature Genetics.

mmSuSiE is an extension of SuSiE, according to Nature Genetics.

It was specifically designed for mixed-model analysis, according to Nature Genetics.

Its purpose is to identify the environmental factors driving the GxE interactions detected by fastGxE, according to Nature Genetics.

The fastGxE Overview describes the same role: mmSuSiE identifies the environmental factors driving detected GxE interactions, according to the fastGxE Overview.

The pipeline also employs the stratified Wald test to visualize and support those interactions, according to the fastGxE Overview.

Conceptually, the workflow moves from recognition of SV-like GxE signals under different contexts to clustering and reconstruction, but for GxE the steps are detection followed by environmental-factor dissection, according to the fastGxE Overview.

Progressive modeling plus local follow-up supports robust reconstruction of the interaction signal, according to the fastGxE Overview.

Speed improvement of 32.98–126.49 times

fastGxE ensures scalability for genome-wide GxE analysis in large biobank studies, according to Nature Genetics.

It achieved speed improvements of 32.98–126.49 times over existing approaches, according to Nature Genetics.

The preprint record reports the same 32.98–126.49 times range, according to the preprint record on PMC.

That range reflects comparisons across existing approaches and settings evaluated in the study, according to Nature Genetics.

The gain matters because prior multi-environment tools were computationally intensive and restricted to candidate regions rather than genome-wide scans, according to Nature Genetics.

Scalability is therefore framed as a requirement for biobank-scale datasets, not merely a convenience, according to Nature Genetics.

UK Biobank test: 32 traits and 67 biomarkers

The real-data demonstration used 32 physical traits from the UK Biobank, according to Nature Genetics.

It also used 67 blood biomarkers from the UK Biobank, according to Nature Genetics.

That combined set covers physical measurements and blood-based measures in the same biobank resource, according to Nature Genetics.

In that application, fastGxE identified 7 genomic loci associated with physical traits, according to Nature Genetics.

Four of those 7 physical-trait loci were novel, according to Nature Genetics.

The method also identified 18 genomic loci associated with blood biomarkers, according to Nature Genetics.

Twelve of those 18 biomarker loci were novel, according to Nature Genetics.

The preprint had reported interim counts of nine physical-trait loci including six novel ones and 26 biomarker loci including 19 novel ones, according to the preprint record on PMC.

The peer-reviewed Nature Genetics figures supersede those preprint counts for the published analysis, with 7 and 18 total loci respectively, according to Nature Genetics.

The new discoveries highlight the dynamic interplay between genetics and the environment, according to Nature Genetics.

They uncover potentially clinically relevant pathways, according to Nature Genetics.

Those pathways could inform personalized interventions and treatment strategies, according to Nature Genetics.

The preprint record uses closely related language, saying the discoveries could inform personalized interventions and treatment strategies and uncover potentially clinically significant pathways, according to the preprint record on PMC.

What the loci mean for personalized strategies

Context-specific genetic effects mean that a variant’s influence can differ across environmental exposures, according to Nature Genetics.

Detecting those differences refines understanding of disease etiology beyond main effects alone, according to Nature Genetics.

Improved etiologic detail can support more accurate models for genetic risk prediction that incorporate GxE effects, according to Nature Genetics.

More accurate risk models are a step toward personalized medicine strategies that integrate both genetic and environmental factors, according to Nature Genetics.

The newly found loci therefore point to dynamic gene–environment interplay rather than fixed genetic determinism, according to Nature Genetics.

The authors frame that interplay as potentially clinically relevant, according to Nature Genetics.

Clinical relevance here refers to pathways that could inform personalized interventions and treatment strategies, according to Nature Genetics.

No specific intervention, dosage, or treatment recommendation is stated in the Nature Genetics abstract, and none is reported here, according to Nature Genetics.

Open-source C++ tool and tutorial

fastGxE is implemented as an open-source C++ package, according to the fastGxE Overview.

It is freely available on GitHub, according to the fastGxE Overview.

The GitHub repository is named chaoning/fastGxE, according to the GitHub repository record.

The tutorial assumes the executable is on the PATH as fastgxe, according to the fastGxE Overview documentation.

If the binary was built from source and is not on the PATH, the tutorial directs users to replace fastgxe with the full path such as ../build/fastgxe, according to the fastGxE Overview documentation.

The software reads PLINK binary genotype data through a prefix passed to --bfile, according to the fastGxE Overview documentation.

For example, --bfile test means the program will use test.bed, test.bim, and test.fam, according to the fastGxE Overview documentation.

The phenotype table passed to --data must satisfy specific rules, according to the fastGxE Overview documentation.

A header row is required, according to the fastGxE Overview documentation.

Fields should be separated by blanks or tabs, according to the fastGxE Overview documentation.

The first column must contain individual IDs, according to the fastGxE Overview documentation.

Only individuals present in both the phenotype table and the genetic relationship matrix will be retained, according to the fastGxE Overview documentation.

Duplicate sample IDs are not allowed, according to the fastGxE Overview documentation.

Typical column usage is --trait pheno for the phenotype, according to the fastGxE Overview documentation.

Typical usage also includes --covar age for continuous adjustment covariates, according to the fastGxE Overview documentation.

It includes --class sex for categorical covariates, according to the fastGxE Overview documentation.

It includes --env-int E1:E40 for interacting environments, according to the fastGxE Overview documentation.

Variables listed in --env-int should not also be listed in --covar or --class, according to the fastGxE Overview documentation.

Interacting environments are standardized by default before GxE testing, according to the fastGxE Overview documentation.

Users can keep the raw scale with --no-standardize-env, according to the fastGxE Overview documentation.

Missing phenotype or covariate tokens default to NA, Na, NAN, NaN, nan, and related variants plus N/A and n/a, according to the fastGxE Overview documentation.

Users can override or extend that missing-value list with --missing-data, according to the fastGxE Overview documentation.

Downstream --test-main and --test-gxe workflows rely on grouped genetic relationship matrix files, according to the fastGxE Overview documentation.

Those files are created with commands such as fastgxe --make-grm --bfile test --out output/test, according to the fastGxE Overview documentation.

For large datasets, the matrix can be computed in partitions with --npart and then merged, according to the fastGxE Overview documentation.

Grouping of related samples uses fastgxe --process-grm --group --grm output/test --cut-value 0.05, according to the fastGxE Overview documentation.

Questions can be left on GitHub issues or sent to [email protected], according to the fastGxE Overview.

Why biobank-scale GxE matters now

The availability of biobank data created the opportunity to test interactions genome-wide rather than in isolated candidate regions, according to Nature Genetics.

Prior single-environment tests lacked power even with large samples, according to Nature Genetics.

Prior multi-environment tests gained power but lost robustness when few environments mattered, according to Nature Genetics.

They also lacked genome-wide scalability, according to Nature Genetics.

They omitted background polygenic structure and heterogeneous noise, according to Nature Genetics.

Those omissions risk false positives and biased architecture estimates, according to Nature Genetics.

fastGxE was built to retain multi-environment power while staying robust to how many environments actually interact, according to Nature Genetics.

It was also built to control background polygenic and interaction effects while modeling heterogeneous noise, according to Nature Genetics.

It was also built to scale to genome-wide analysis in large biobanks, according to Nature Genetics.

The reported 32.98–126.49 times speedup directly addresses the computational barrier, according to Nature Genetics.

The UK Biobank results on 32 physical traits and 67 blood biomarkers demonstrate that scaled testing can surface novel loci, according to Nature Genetics.

With 4 novel physical-trait loci and 12 novel biomarker loci, the majority of biomarker discoveries in this scan were new in this interaction context, according to Nature Genetics.

That pattern underscores how much context-dependent signal remains hidden when only main effects are tested, according to Nature Genetics.

Sources referenced