Leveraging Novel Statistical Methods And Big Data To Improve Imaging Genetics

Abstract

Overall Abstract Background The brain is a highly heritable organ with obvious associations to psychiatric diseases. Despite initial hype at the beginning of this century, it has been difficult to find brain related genotype-phenotype associations that replicate in independent samples. However, with the recent emergence of large consortia such as ENIGMA and ABCD, harmonized processing and analysis protocols across multiple sites allows for unprecedented power to detect genetically determined brain variation. In addition, novel statistical methods for big data, customized to the polygenic nature of brain phenotypes, have increased the potential to quantify variance due to genetic factors, and to identify genetic loci that replicate across GWAS samples. Here, we present the potential of combining large scale brain imaging data collection with novel biostatistical tools, and provide an update of the field related to the following topics: 1) Novel statistical approaches for imaging genetics (Thompson et al). This talk will propose novel statistical approaches for imaging genetics data derived from the field of Bayesian psychometrics. This talk will also cover how these methods will be applied to the Adolescent Brain and Cognitive Development (ABCD) study, a nationwide 19 site NIH funded study that will recruit over 10,000 children aged 10-11 and followed for 10 years, with genetics, a twin component, and brain imaging every other year. The ABCD repository, which will be released to the public beginning December 1, 2017 and which will be of great value for imaging genetics studies, will be described in detail. The speaker, Dr. Thompson, is the Director of Biostatistics for the ABCD Consortium. 2) ENIGMA update - discoveries of novel brain associated gene variants (Hibar et al). Results from most recent GWAS of brain structure provides novel insight into common variants associated with cortical thickness and surface area using meta-analytic framework in more than 20,000 participants. Novel genome-wide significant variants were observed in the lateral occipital and pericalcerine gyri. These results provide insights into the common genetic architecture of the brain in the largest effort to date. 3) Bayesian approach to PGC - ENIGMA data reveals overlapping variants between schizophrenia and brain structure volumes (Smeland et al). Combined analysis of GWAS data using conditional false discovery rate methods provides increased power to detect overlapping genetic loci. Using this approach, we here identified genetic loci shared between schizophrenia and volumes of hippocampus, putamen and intracranial volume, suggesting novel molecular genetic mechanisms. 4) Imaging endophenotype revisit (Fan et al). Although neurodevelopmental processes seldom occur in discretized sets, imaging genetic analyses often rely on measures of morphologically defined region of interests (ROI). However, arbitrarily defined ROI can sometimes introduce bias. With larger training data available, it is now possible to extract reliable imaging features pertaining to genetic variations other than traditionally defined ROI. Here we demonstrate recent studies on learning new endophenotypes and the yield for discovering genetic influences on neuroanatomy. The learnt manifolds of anatomical structure can either be used for correcting the bias, e.g. population structure, or improving the power to detect genetic effects. This is important for extending imaging genetics research beyond caucasian samples.