S7 Fig -

<p>Comparison of training data effective sample sizes <b>(A)</b> and total samples (B) for disease and quantitative traits, among Comparator, Standard and Enhanced PRS. Total training sample sizes for the Enhanced PRS, Standard PRS, and the best-performing comparator PRS, for disease traits (<b>A</b>) and quantitative traits (<b>B</b>) [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0307270#pone.0307270.ref020" target="_blank">20</a>]. For disease traits, the x-axis is the effective sample size, defined as 4 / ((1/n<sub>0</sub>) + (1/n<sub>1</sub>)), where n<sub>0</sub> is the total number of controls, and n<sub>1</sub> is the total number of cases. Where training data came from meta-analysis of multiple GWASs, the total numbers are used, as it was not always possible to obtain accurate per-GWAS numbers for the PGS Catalog PRSs. A dashed connecting line is used where the comparator PRS sample size is larger than the Enhanced PRS sample size. Only those traits for which at least one comparator PRS was available are displayed. Traits with overlapping samples in the Standard/Enhanced PRS training are excluded. In addition, ischaemic stroke is not shown, because the best comparator PRS (Abraham et al 2019, doi: <a href="https://doi.org/10.1038/s41467-019-13848-1" target="_blank">10.1038/s41467-019-13848-1</a>) was trained using a complex combination of PRSs for 19 different diseases and quantitative traits; hypertension is also excluded because the only comparator PRS (Said et al 2018, doi: <a href="https://doi.org/10.1001/jamacardio.2018.1717" target="_blank">10.1001/jamacardio.2018.1717</a>) used a combination of disease and quantitative trait data. Refer to <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0307270#pone.0307270.g001" target="_blank">Fig 1</a> legend for disease and quantitative trait abbreviations.</p> <p>(PDF)</p>