Using local ancestry inference to improve schizophrenia polygenic risk score prediction in admixed populations

Abstract

Schizophrenia is a severe mental disease with a prevalence of up to 1% in the world population and one of the main causes of disability in developed countries. Schizophrenia is a multifactorial disease and heritability is around 0,8. This suggests that the genome contains a large amount of information that has the potential to be used as genetic markers for diagnosing these diseases. In the last years, the expansion of consortia in psychiatric genetics has allowed the creation of a polygenic risk score for schizophrenia (PRS) based on the results of genome-wide association studies (GWAS). One of the limitations of the PRS is that it relies on GWAS results, which are mainly composed by Caucasian/European population, making the translation to admixed populations, as Brazilian, a difficult task. Previously, the group carried out several tests to verify which is the best model to use PRS in the Brazilian sample. In general, we demonstrate that the prediction is less than what is expected for European samples. The objectives of this proposal are: 1) Verify the impact of admixture, measured by local ancestry, on the polygenic risk score (PRS) for schizophrenia; 2) Check if there is an improvement in the PRS prediction after the decomposition of the polygenic risk score in three subpopulations (European, African and Native American) for each individual. For the analysis of the PRS, we will use the group's already genotyped cohorts (N = 2850) with the Global Screening Array. The analysis will be run entirely in UNIX operating systems using the command line language as well as the statistical program R, PLINK and PRSice. First, the student will break down the genome of each of the 2850 individuals into chromosomal segments for the three most frequent populations in the Brazilians, European, African and Native American ancestries. For that, we will use the Beagle tool for the genomic phasing of the genotyping data and then we will use the XGmix program to decompose into three populations. After the decomposition, we will use the Tractor tool to run the schizophrenia GWAS in the 3 populations and then check the increase in genetic prediction when the individual genome is decomposed. As the main finding, we intend to demonstrate an effective solution for the use of the polygenic risk score in an admixed population, taking into account the individual admixture pattern.