Variant-level odds ratios for 18 phenotypes

This dataset is from the paper "Extracting and calibrating evidence of variant pathogenicity from population biobank data," which evaluates the broad potential of using population cohort data as evidence of pathogenicity in variant assessment. The dataset contains variant-level odds ratios in 18 phenotypes spanning 41 genes, calculated using case data from 469,803 UK Biobank participants. These odds ratios can be used as PS4 (population) evidence of pathogenicity within the ACMG/AMP variant interpretation framework using the calibrated thresholds provided in Table 2 (for specific phenotypes), Supplementary Table 5 (for specific genes), or Table 3 (overall).medRxiv preprint: https://www.medrxiv.org/content/10.1101/2024.08.14.24311911v2We share odds ratio data by phenotype, yielding 18 files. Each file has the following columns:<b>gene:</b> Gene in which the variant is located (e.g., <i>LDLR</i>)<b>variant:</b> Variant in chr-pos-ref-alt format (e.g., 11-108227661-C-T)<b>odds_estimate:</b> Odds ratio that measures disease (i.e., phenotype) enrichment for the variant<b>odds_lower:</b> Odds ratio 95% confidence lower bound<b>odds_upper:</b> odds ratio 95% confidence upper bound<b>cases:</b> Case data used in part of the odds ratio calculation<br>(e.g., 1|0|1|0|0, meaning 2 participants with the disease and 3 without)<b>revel_score:</b> REVEL score for the variant, obtained via the Variant Effect Predictor (VEP)<b>clinvar_status:</b> Variant ClinVar status (one of "P/LP", "B/LB", "VUS", or "Absent")Note that the underlying case data used to calculate odds ratios (i.e., the "cases" column) shouldn't always be treated as case data in the traditional sense. For some phenotypes, a "1" or "0" represents a proxy for the disease. For example, for familial hypercholesterolemia, adjusted LDL-C ≥ 190 mg/dL is used as a proxy for an FH case. See the "Clinical endpoints and endophenotypes" section of the methods for more information.Additionally, not all variants listed have associated odds ratios. Typically comprising the majority of variants, this is because for many variants, all participants with the variant don't have the associated disease. If there are not enough such participants, a corrected odds ratio will be a distorted measure. See the "Calculating variant-level odds ratios from population cohort data" section of the methods for more information.