We conduct a genome-wide association study (GWAS) of educational attainment (EA) in a sample of ~3 million individuals and identify 3,952 approximately uncorrelated genome-wide-significant single-nucleotide polymorphisms (SNPs). A genome-wide polygenic predictor, or polygenic index (PGI), explains 12–16% of EA variance and contributes to risk prediction for ten diseases. Direct effects (i.e., controlling for parental PGIs) explain roughly half the PGI’s magnitude of association with EA and other phenotypes. The correlation between mate-pair PGIs is far too large to be consistent with phenotypic assortment alone, implying additional assortment on PGI-associated factors. In an additional GWAS of dominance deviations from the additive model, we identify no genome-wide-significant SNPs, and a separate X-chromosome additive GWAS identifies 57.

Polygenic indexes (PGIs) are DNA-based predictors. Their value for research in many scientific disciplines is rapidly growing. As a resource for researchers, we used a consistent methodology to construct PGIs for 47 phenotypes in 11 datasets. To maximize the PGIs' prediction accuracies, we constructed them using genome-wide association studies-some of which are novel-from multiple data sources, including 23andMe and UK Biobank. We present a theoretical framework to help interpret analyses involving PGIs. A key insight is that a PGI can be understood as an unbiased but noisy measure of a latent variable we call the "additive SNP factor." Regressions in which the true regressor is the additive SNP factor but the PGI is used as its proxy therefore suffer from errors-in-variables bias. We derive an estimator that corrects for the bias, illustrate the correction, and make a Python tool for implementing it publicly available.

Distinguishing between damaging and neutral missense variants is an ongoing challenge in human genetics, with profound implications for clinical diagnosis, genetic studies and protein engineering. Recently, deep-learning models have achieved state-of-the-art performance in classifying variants as pathogenic or benign. However, these models are currently unable to provide predictions over all missense variants, either because of dependency on close protein homologs or due to software limitations. Here we leveraged ESM1b, a 650M-parameter protein language model, to predict the functional impact of human coding variation at scale. To overcome existing technical limitations, we developed a modified ESM1b workflow and functionalized, for the first time, all proteins in the human genome, resulting in predictions for all ~450M possible missense variant effects. ESM1b was able to distinguish between pathogenic and benign variants across ~150K variants annotated in ClinVar and HGMD, outperforming existing state-of-the-art methods. ESM1b also exceeded the state of the art at predicting the experimental results of deep mutational scans. We further annotated ~2M variants across ~9K alternatively-spliced genes as damaging in certain protein isoforms while neutral in others, demonstrating the importance of considering all isoforms when functionalizing variant effects. The complete catalog of variant effect predictions is available at: https://huggingface.co/spaces/ntranoslab/esm_variants.

Predicting the effects of coding variants is a major challenge. While recent deep-learning models have improved variant effect prediction accuracy, they cannot analyze all coding variants due to dependency on close homologs or software limitations. Here we developed a workflow using ESM1b, a 650-million-parameter protein language model, to predict all ~450 million possible missense variant effects in the human genome, and made all predictions available on a web portal. ESM1b outperformed existing methods in classifying ~150,000 ClinVar/HGMD missense variants as pathogenic or benign and predicting measurements across 28 deep mutational scan datasets. We further annotated ~2 million variants as damaging only in specific protein isoforms, demonstrating the importance of considering all isoforms when predicting variant effects. Our approach also generalizes to more complex coding variants such as in-frame indels and stop-gains. Together, these results establish protein language models as an effective, accurate and general approach to predicting variant effects.