Assessment of the impact of genetic variability in human accelerated regions
As regards the genetic etiology of schizophrenia (SZ), recent studies have suggested the relevance of gene expression regulatory mechanisms that modulate the genetic effects on the development and plasticity of the human brain.
Different hypotheses have indicated the evolutionary component of schizophrenia, all converging in the suggestion that SZ emerged as a costly trade-off in the evolution of ontogenetic mechanisms underlying human-specific brain development and, finally, complex cognitive abilities. Human accelerated regions (HARs) are evolutionarily conserved genomic regions that have experienced significant changes after humans' divergence from chimpanzees. This accelerated divergence of HARs is suggested to reflect their role in human evolution and their association with some human-specific traits. If some HARs regulate human-specific neurodevelopment mechanisms, then sequence variants in these regions would likely impact on disorders such as SZ. Very few studies have investigated the role of HARs in schizophrenia. One study showed that HARassociated schizophrenia genes were enriched for processes involved in synaptic formation. Another suggested that the detected enrichment could mostly be accounted for by linkage disequilibrium (between HARs and known genomic categories). Further studies on the role of HARs in schizophrenia are needed and they would benefit from the use of more specific brain phenotypes.
We plan to apply different machine learning algorithms to explore the role of genetic variability in HARs in brain structural differences observed in patients with SZ compared to healthy subjects. We hypothesize that the presence/intensity of structural abnormalities in the brains of SZ patients will be, to some extent, explained by differences in the relative frequencies of genetic variants in HARs (and interactions). This will be developed in a discovery sample of 150 SZ patients and 150 healthy controls previously recruited, all with structural T1 images and available blood/saliva samples. In addition, a replication sample of 80 SZ patients and of 80 controls will be also available. Individual T1 images will be transformed to voxel-based morphometry (VBM) maps. The samples will be genotyped using the Infinium PsychArray (Illumina). All the genotyped Single Nucleotide Polymorphisms (SNPs) will be classified according to whether they fell inside or outside any HAR, and, subsequently, an LD-weighted score will be generated. The potential effect of HARs variability will be evaluated following three steps. 1) Regularized Lasso regression will be used to select sub-lists of HARs variants with relevant additive effects on brain structural differences between patients and controls. 2) Taking the selected HARs variants from step1, regression tree algorithms will be fit to fully explore the potential epistatic effects. 3) Regression analyses will be performed by considering the lists of HARs variants identified in step1 and 2 as independent variables: 3.1) logistic regression with group label as the dependent variable; 3.2) linear regression with level of structural abnormality as the dependent variable. As a final output, a brain map including a list of HARs and interactions involved in brain structural abnormalities in SZ will be reported.