Our research project covers projects in statistical genetics with a focus on brain image-derived phenotypes (IDPs) in UK Biobank. The goal is to address the following research gaps: (i) the need for an updated pipeline for newly released brain imaging data; (ii) the inclusion of minorities in UKB genetic analysis, including but not limited to participants with non-white British ancestry and with mismatched genetic and reported sex; (iii) the development of robust meta-analyses for overlapping genetic studies.
On the applied side, we will develop an update to the pipeline for preprocessing brain imaging genetics for UK Biobank. This update will include the Y chromosome, which has not been studied in main genetic scans of all UKB IDPs due to loss-of-Y (some older males are missing Y chromosomes), and lack of linkage disequilibrium (there is only one copy of the Y chromosome, so it does not recombine). This update will also improve X chromosome analysis by modeling sex differences in minor allele frequencies (sd-MAF), observed in large consortia such as the Thousand Genome Project. This will lead to a new and improved version of UKB brain imaging summary statistics (previously published as BIG40).
On the theoretical side, we will develop robust statistical methods that account for multiple ancestry and sex-linked variants and phenotypes. Non-white British samples have been excluded in earlier brain imaging studies due to their small sample size. However, the transferability of previous GWAS results to minority populations has not been examined. Inclusive studies should include all UKB participants, especially those excluded due to sex mismatch. Studies have shown that most samples with sex mismatch are transgendered, which further demands for transgender-inclusive genome-wide association methods.
Our new methods will also correct bias in meta-analysis arising from overlaps between studies. When data linkage is not possible between consortia, participants may be included in multiple studies, and we may not know which participants overlap. This leads to bias in meta-analysis p-values. We will find bounds for the bias, and use UKB to demonstrate this theory. This will improve the quality of summary statistics when consortia such as People of the British Isles or the 100,000 Genomes Project are combined with UKB.
Overall, the work described in this proposal will improve the use of UKB as a cohort for reference and discovery by improving the quality of brain imaging genetics, and the transferability of genetic findings using robust statistics.