Like the figures folder, this folder contains jupyter notebooks which generated the supplementary figures for the manuscript. As the same caveat with respect to individual-level UK Biobank data holds here, it is similarly not possible to use these notebooks as-is until other analyses are re-run locally. They are provided as-is for replication and future analysis.
With few exceptions, there is a one-to-one correspondence between notebooks and supplementary figures, tables, and data. Exceptions are listed below:
- Data for Table S1 are derived from the notebook for Figure S1
- The "Figure S1b" notebook results are not present in the manuscript version of the supplement
- Figure S3 was derived using the phenotype pipeline from the scripts directory
- Visualizations for Data S2 and S3 (phenotype and gene contribution scores) were respectively made using the notebook for Figure S1 and
gene_plots.py. Numeric values for these contribution scores are available for download on the Global Biobank Engine
Results from profiling the predictive performance of polygenic risk scores (PRS), including the DeGAs polygenic risk score (dPRS), are in the results folder. Each file contains results from either PRS or dPRS in one of the four analysis populations: train, dev, and test sets of white British individuals, and an additional test set of non-British white individuals (nbw) from UK Biobank.
Columns in these files are as follow:
pheno: Global Engine Phenotype code (see reference folder for more information).n: Number of cases, or individuals with quantitative measurementbeta2: Regression coefficient for being in the top 2% of PRS or dPRS for a given trait (adjusted for covariates)auc: Area under the receiver operating curve using PRS or dPRS (unadjusted) as the classifying score, for binary traitspearsonr: Pearson's r between PRS and dPRS and quantitative trait valuesspearmanr: Spearman's rho (rank correlation) between PRS and dPRS and trait values
Additional information on these measures, including covariate adjustment, can be found in the Methods section of our manuscript.
We also computed DeGAs on an additional publicly available dataset of summary statistics from UK Biobank GWAS. Please refer to the documentation in the extra folder for more details.