Identification of predictive models for psychiatric disorders from genetic variants

• Grant number: 19/10409-0
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: October 01, 2019
• End date: September 30, 2020
• Field of knowledge: Health Sciences - Medicine - Psychiatry
• Principal Investigator: Síntia Iole Nogueira Belangero
• Grantee: Tamiris Vieira da Fonseca
• Host Institution: Escola Paulista de Medicina (EPM). Universidade Federal de São Paulo (UNIFESP). Campus São Paulo. São Paulo , SP, Brazil

Abstract

Psychiatric Disorders (PD) are severe public health problems. These disorders are highly heritable and have a complex, overlapping polygenic architecture. Besides, they include a combination of genetic and environmental factors. Because of this, it is important to examine shared genetic risk factors for polygenic prediction of PD. Aim: This study aims to identify patterns among psychiatric disorders based on genetic variables and test predictive models for diagnosis of PD in a large sample of Brazilians. Materials and Methods: We will evaluate 3.497 subjects, distributed among case and control groups, from different cohorts with different PD diagnoses, such as schizophrenia, first psychotic episode, posttraumatic stress disorder, among others. For genotyping, we used the Human OmniExpress Beadchip, the Infinium PsychArray BeadChip and the Infinium Global Screening Array BeadChip. We will perform a genomic imputation to compile and homogenize the genotyping data of the different microarrays. An exploratory analysis will be carried out using the non-supervised machine learning methodology to observe how genetic variables previously associated with PDs are classified in the Brazilian population. The predictive analysis will test two PD diagnostic predictive models: 1) supervised machine learning model; and 2) Polygenic Risk Score (PRS). The machine learning analyses will use genetic polymorphisms that have been already previously associated with PT in a large association study (GWAS) carried out by the Cross-Disorder Consortium group of the International Consortium of Psychiatric Genomics (PGC) as predictors. All analyses will be carried out on UNIX systems using the following tools: software R, PRSice and PLINK. Expected Results: We intend to find a model that can predict diagnoses or groups of PT diagnoses using computational tools based on the genetic variants in a Brazilian sample and, in addition, verify whether the initial diagnostic classification of the subjects reflects their genetic basis. (AU)