M19 - ESTIMATE SPATIOTEMPORAL IMPACT OF BRAIN VARIANTS

Abstract

Background Functional characterization for genetic variants is a major challenge in whole-genome sequencing-based studies. Recent approaches, such as TiSAn (Vervier et al., 2017) or GenoSkyline (Lu et al., 2016), estimate tissue-specific impact of variations, in particular in human brain tissues. However, such annotations do not provide insights on which brain regions or development time points might be especially vulnerable to a given variant. Methods In this work, we propose to integrate spatiotemporal gene expression from BrainSpan (BrainSpan, 2014) to estimate what the 'context matrix' of a variant is. Variants found in non-coding regions are represented as a combination of gene expression matrices, where weights are based on associations demonstrated in SLINGER models (Vervier et al., 2016). Results We validate our approach on psychiatric disorder datasets, and use temporal patterns to discriminate early- and late-onset damaging variants. Brain region-specific variants also help to identify combined mechanisms of action, in complex traits. Discussion Spatio-temporal profiles could also be combined with polygenic risk score approaches, and provide new dimensions to study psychiatric disorders.