September 3, 2026
Joey Poomarin Phloyphisut, Software Engineer, and Cory McLean, Senior Staff Software Engineer, Google Research
We evaluate ways to improve cross-population genetic risk prediction and find that while transfer learning from European cohorts improves prediction in small underrepresented populations, it degrades accuracy once target cohort sample sizes grow, especially for traits with population-specific genetic architectures.
Polygenic risk scores (PRSs) are used to predict disease risk from genetic variants. They usually incorporate the influence of hundreds to millions of genetic variants. However, their adoption for clinical decision making is currently low, partly because historical genome-wide association studies (GWASs) have overwhelmingly evaluated European cohorts, resulting in severe accuracy drops when applied to non-European populations. These accuracy differences arise because of cross-population differences in genetic architectures, population structure, and variant allele frequencies.
Moreover, conducting de novo GWASs across hundreds of thousands of individuals is cost-prohibitive for most healthcare systems. Transfer learning from existing European-centric GWAS, augmenting these large-scale results with target-population-specific GWAS, provides a potential solution.
To that end, in this blog post we describe a study evaluating PRS performance in a target non-European population while varying the size of both the target population and European populations used to create the predictive model across eight clinical traits. Specifically, we evaluate transferability of PRS ascertained in hundreds of thousands of European individuals within the UK Biobank (UKB) to samples within Biobank Japan (BBJ), a deeply-phenotyped cohort of nearly 200 thousand Japanese individuals. Our primary objective is to provide systematic, empirical guidelines on how cross-population GWAS and PRS model training should be performed to optimize predictive performance in a target population.
The presence of two large-scale, deeply genotyped and phenotyped datasets (UKB and BBJ) enables dataset ablation experiments to systematically evaluate PRS performance as a function of sample size. We selected eight clinically relevant traits measured in both populations for evaluation: body mass index (BMI), systolic blood pressure, diastolic blood pressure, red blood cell count, white blood cell count, high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), and blood glucose. In UKB, the estimated fraction of trait variance explained by the measured genetic variants (single nucleotide polymorphism heritability) ranged from 0.07–0.28.
Three methods were used to evaluate the PRS performance transferability:
In all experiments, the PRS models were evaluated on the same held-out set of BBJ samples.
For each experiment, we quantified model performance using Pearson correlation. As expected, European discovery data provides a helpful baseline when target population data is limited or non-existent. However, the target-population-specific models outperform at sample sizes of 15k or more, as it appears that co-training with external European data limits the target population accuracy gains achieved through higher sampling.
This surprising observation held across all phenotypes we examined. When the target sample size is extremely small (e.g., 5,000 samples), pooling European UKB data during training provides a valuable statistical boost. As the target sample size used to train the PRS model increases, the benefit of out-of-population pooling diminishes.
While the trend holds across phenotypes, the crossover point at which the success of using European data diminishes depends heavily on the specific trait being studied. We can quantify the extent of “shared genetics” using genetic correlation of the same trait across populations. We observe that “conserved” traits, or those with higher genetic correlations between the two populations, retain the benefits of pooling UKB European training data up to much larger target sample sizes (25-40k+ samples) before target-population-specific training data reaches parity.
In sharp contrast, lipid levels (HDL, LDL) and blood glucose exhibit much smaller BBJ sample sizes beyond which optimal UKB sample sizes are smaller than maximum, and that optimal sample size is also smaller. These highly population-specific traits do not benefit as much from UKB data, since that data is farther out-of-distribution.
The above experiments restricted PRS model input variants to those discovered in the UKB European population, excluding any trait-associated variants unique to BBJ samples. To extend the analyses to capture these, we created two additional prediction methods.
First, we ran GWAS on each BBJ sample size. The first new method performed a cross-population meta-analysis using the full UKB GWAS and the sample-size-specific BBJ GWAS to identify candidate variants, and then fit an elastic net on those variants. The second method used PRS-CSx to combine the two sets of GWAS summary statistics.
By tracking the net performance gain of meta-analysis and PRS-CSx across varying discovery sample sizes, we observed differences in performance across BBJ sample sizes.
The influence of meta-analysis is low for conserved traits, largely due to reduced statistical power in the much smaller BBJ GWAS sample sizes. However, for population-specific traits like HDL and LDL, and to a lesser extent blood glucose, meta-analysis substantially outperforms single-population discovery. The gains are primarily due to modifying the elastic net variant input: including UKB European samples during training slightly improved prediction when using 10,000 or fewer BBJ samples. This improvement was not seen with larger BBJ samples.
Because PRS-CSx dynamically weights population-specific models, its performance is theoretically less sensitive to conserved vs population-specific traits. However, we observed that the model requires more data than elastic net models to perform well. For target sample sizes under 25k, PRS-CSx performs worse than the strongest corresponding elastic net model in all phenotypes except BMI. As sample sizes approached 100k, PRS-CSx matched or exceeded the best performing model across all phenotypes except blood glucose.
Systematic evaluation of cross-population genomic prediction reveals that larger out-of-population datasets are not always beneficial when applying polygenic risk scores (PRSs) to underrepresented ancestries. Specifically, while transfer learning from the large European cohort of UK Biobank provided a statistical boost at low target population sizes (under 15,000 samples in Biobank Japan), it actually degraded predictive accuracy as the BBJ sample size grew. This performance crossover is trait-dependent: genetically conserved traits (such as BMI) retain the benefits of external data pooling up to larger sample sizes, whereas population-specific traits (such as lipids and blood glucose) show diminished benefits with fewer samples. Importantly, advanced multi-ancestry methods like PRS-CSx require substantial target-population samples to outperform simpler approaches, while cross-population meta-analysis offers robust benefits for population-specific traits at smaller sizes. Ultimately, these findings emphasize that optimizing predictive performance in diverse populations will require both the expansion of local, diverse biobanks and careful selection of modeling strategy that is tailored to both trait heritability and sample size.
We sincerely thank Biobank Japan and our collaborators at RIKEN and The Institute of Medical Science The University of Tokyo for enabling this research, as well as additional Google collaborators: Babak Behsaz, Andrew Carroll, Farhad Hormozdiari, and Taedong Yun. Our thanks also go to Hiroki Kayama and Joe Ledsam for institutional support, and Michael Brenner and Katherine Chou for their leadership support.
Facts Only
* The study evaluated PRS performance across eight clinical traits: BMI, systolic blood pressure, diastolic blood pressure, red blood cell count, white blood cell count, HDL cholesterol, LDL cholesterol, and blood glucose.
* The evaluation used data from the UK Biobank (UKB) and Biobank Japan (BBJ).
* In UKB, the estimated fraction of trait variance explained by genetic variants ranged from 0.07–0.28.
* PRS models were evaluated on a held-out set of BBJ samples.
* For small target sample sizes (e.g., 5,000 samples), pooling European UKB data provided a statistical boost.
* As the target sample size used to train the PRS model increased, the benefit of out-of-population pooling diminished for most phenotypes.
* For conserved traits (e.g., BMI), the benefits of pooling external data were retained up to larger target sample sizes (25-40k+ samples).
* Lipid levels (HDL, LDL) and blood glucose exhibited smaller optimal UKB sample sizes compared to other traits.
* Cross-population meta-analysis showed low influence for conserved traits but substantial performance gains for population-specific traits like HDL and LDL at smaller BBJ sizes.
* PRS-CSx required more data than elastic net models to perform well for target sample sizes under 25k, except for BMI.
Executive Summary
Transfer learning from European cohorts improves prediction in small, underrepresented populations but degrades accuracy as target cohort sample sizes increase, particularly for traits with population-specific genetic architectures. This phenomenon occurs because historical Genome-Wide Association Studies (GWASs) are overwhelmingly based on European cohorts, leading to differences in genetic architectures and allele frequencies when applied to non-European populations. The study evaluated Polygenic Risk Scores (PRSs) across eight clinical traits using data from the UK Biobank (UKB) and Biobank Japan (BBJ).
The results show that pooling European data provides a statistical boost when target population sample sizes are small, such as below 15,000 samples. However, for genetically conserved traits like BMI, the benefit of external data pooling persists up to larger target sizes. Conversely, highly population-specific traits, such as lipid levels and blood glucose, showed diminished benefits from UKB data as BBJ sample sizes increased.
The research explored methods to mitigate these differences, including cross-population meta-analysis and PRS-CSx. These advanced methods demonstrated that their performance is sensitive to the trait type; they provided substantial gains for population-specific traits at smaller sample sizes through modification of input variant weighting. Ultimately, optimizing predictive performance requires tailoring the modeling strategy to account for both trait heritability and the specific genetic correlations between populations.
Full Take
The core tension revealed is the trade-off between generalization from large external datasets and specificity derived from local population structure when applying genetic prediction tools. The finding that transfer learning degrades accuracy as the target cohort grows suggests a fundamental limitation in assuming shared genetic architecture across populations, even for traits considered "conserved." This challenges the default assumption that larger sample sizes inherently yield better predictive power regardless of underlying genetic covariance.
The divergence observed between conserved traits (retaining benefits up to 40k samples) and population-specific traits (diminishing benefits quickly) highlights a critical pattern: the utility of external data is modulated by the relationship between the trait's heritability and its specific genetic correlation structure across ancestries. For highly homogeneous traits, the signal from distant cohorts remains robust; for highly divergent traits, the localized, fine-grained information from the target population becomes overwhelmingly dominant as sample sizes expand.
The necessity of advanced methods like PRS-CSx—which dynamically weight source data based on trait covariance—suggests that a one-size-fits-all transfer learning approach is insufficient. This points toward an implication that optimizing cross-population prediction requires moving beyond simple data aggregation to explicitly model the heterogeneity inherent in genetic architectures, necessitating both expanding local data collection and adaptive statistical modeling tailored to the specific trait being studied.
Bridge Questions: How can we develop metrics or models to quantify "shared genetics" more robustly than genetic correlation to better predict when transfer learning will degrade? What are the downstream clinical consequences if current clinical guidelines continue to rely on pan-ancestry PRS, potentially misstating risk for population-specific conditions? If optimizing prediction relies on tailoring methods to trait covariance, what institutional and regulatory shifts are required to integrate these nuanced results into clinical decision-making?
Sentinel — Human
This text reads as a detailed, expert-driven scientific communication outlining complex findings from a multi-population genomic study, exhibiting high structural coherence and depth appropriate for peer-reviewed or specialized blog content.