Fig. 3: Results of gene-level burden tests.

A Manhattan plot for gene-level burden tests across all 596 imaging phenotypes. For significant gene-trait pairs appearing in multiple burden models, we plotted the association with the smallest P-value. P-values are derived from two-sided tests based on z-scores (or equivalently a χ² test with one degree of freedom. The dashed lines indicate the gene-level significant P-value threshold at 1 × 10⁻⁹ as well as the suggestive P-value threshold at 1 × 10⁻⁸. The multiple testing thresholds were derived from empirical null distribution with details provided in Method. The y-axis is capped at 1 × 10⁻³⁰ and only gene-trait pairs with P < 1 × 10⁻³ were included. The color legend and x-axis are same as the Manhattan plot for variant-level associations in Fig. 2A. Significant genes were labeled. B Effect sizes for all non-redundant significant gene-trait pairs (n = 67, corresponding to Table 1) within each imaging category. The dashed line indicates an effect size of zero. We labelled all the genes that had the largest positive and negative effects within each imaging category and genes that had an absolute effect size greater than 1.5. C Distribution of total brain volume (in cm³) of mutation carriers (N = 7 individuals) versus non-carriers (N = 40,739 individuals) of PTEN among all the individuals (N = 40,746) included in the exome-wide association test for total brain volume. To illustrate, we select the burden model with the smallest P-value, that is “plof_int5”, which means the aggregation of pLoF variants and missense variants with a minor allele frequency (MAF) cutoff at 1 × 10⁻⁴. In the violin and boxplots shown, the violin plot illustrates the full distribution of total brain volume across carriers and non-carriers of PTEN burden variants using kernel density estimation, with the width of the violin representing data density at each value. Overlaid boxplots summarize key statistics: the center line denotes the median; the bounds of each box represent the first (Q1) and third (Q3) quartiles; and the whiskers extend to the most extreme data points within 1.5 times the interquartile range (IQR) from the box. The black diamonds indicate group medians. D The effect size estimates (circles, triangles, and diamonds at the centers of bars for distinct MAF cutoffs) and the corresponding 95% confidence intervals (bars) across different burden models of PTEN (N = 40,476 individuals). The color indicates distinct MAF cutoffs. The y-axis includes all the burden models (Methods). Only burden models with minor allele count >5 (after aggregation) were tested (Methods) and plotted.