Société Française de Bioinformatique

Mots-Clés

Human Genetics Machine learning neuroimaging psychiatric disorder

Description

Background: A large body of neuroimaging research has documented a variety of abnormalities related to complex neuropsychiatric disorders and psychological traits. Building on these results, the study of neuroimaging phenotypes and their associated genetic and environmental factors has become a central component of ongoing research, with a strong potential for improving our understanding of disease etiologies and identifying promising therapeutic targets1. Genetic studies have arisen as an area of particular interest in this context. Family-based and genome-wide association studies (GWAS) have shown that both mental disorders and neuroanatomical phenotypes are highly heritable2,3, and highlighted strong genetic correlations between neuroimaging phenotypes and mental health outcomes4,5. However, our knowledge about the genetic variations influencing human brain structure and function remains limited. In particular, questions remain on how to best design robust, non-biased descriptors of brain MRI (magnetic resonance imaging) phenotypes to better understand the underlying biological pathways and support the development of biomarkers addressing the lack of gold standards in mental health diagnosis6.

Objectives: The goal of the internship is to conduct pilot analyses to investigate the potential of machine learning approaches to infer latent neuroimaging phenotypes displaying maximum fit with the genetics of mental health outcomes. The project build on readily available data and approaches developed by others and within the team7–9.

Missions: Understand the objectives, constraints and data used for the proposed topic. Assess the suitability and implement existing machine learning tools to achieve the objective, and summarize this work to the team.