THE PGC-PSTD ANCESTRY WORKING GROUP: IMPROVING THE INCLUSION OF ADMIXED INDIVIDUALS IN PSYCHIATRIC GWAS

The latest PGC-PTSD GWAS achieved a considerable increase in the sample sizes of non-European populations, including more than 50,000 African American (AA) and 13,000 Latin American (LA) individuals. One of the difficulties in analyzing these samples is that they are usually very admixed, with different patterns of admixture according to the region they have been collected. The PGC-PTSD Ancestry working group is focused on developing new tools to allow proper inclusion of admixed individuals in GWAS and to facilitate the use of these tools for LMIC collaborators. Previously, we developed Tractor, a gene discovery tool for admixed populations that uses Local Ancestry Inference (LAI) to allow the well-calibrated inclusion of these individuals in GWAS studies. Currently, one of the primary efforts is to comprehensively test strategies for LAI to provide guidelines for parameter setting and reference panel composition. For this, we have simulated admixed individuals considering different patterns of admixture (AA-2 way and LA-3 way), demographic models, software settings, genomic data types, and reference panels. These simulations allow us to define best practice guidelines for researchers to use when analyzing diverse populations to produce the highest accuracy results across ancestries.

In our tests, we observe that Amerindigenous ancestry tracts suffer from notably reduced accuracy as compared to European and African tracts. When miscalls occur, LAI error rates are most frequently in the direction of calling European ancestry in true Amerindigenous sites than other error modes. Though our investigations are directly responsive to realistic admixed Latin American cohort compositions, the trends we characterize are broadly useful to inform best practice for local ancestry inference across diverse admixed populations. In parallel with this initiative of improving LAI analysis accuracy, our working group is developing a friendly pipeline for Tractor so other consortia can apply this approach for large sample sizes. As a future perspective, the PGC-PTSD ancestry working group will also work on the development of new tools for post GWAS approaches as polygenic scores.