Integrated genomics of schizophrenia: finding a pathway to personalised medicine

SZ is a complex psychotic disorder that is currently diagnosed exclusively on clinical presentation. While many variants have been associated with the disorder, the challenge remains to develop diagnostics from this enormous... [ view full abstract ]

SZ is a complex psychotic disorder that is currently diagnosed exclusively on clinical presentation. While many variants have been associated with the disorder, the challenge remains to develop diagnostics from this enormous amount of information. To overcome some of the barriers of complexity, we deploy a multidimensional approach based on the integration of genic risk for both SNPs and CNVs with known association to schizophrenia, with rare variants derived through whole genome sequence (WGS) analysis. Firstly, a logistic regression classifier approach was conducted to account for PRS for both SNPs and CNVs based of the PGC summary statistics. This was conducted on a cohort size of 433 cases, 302 psychiatrically screened controls and 1951 unscreened-controls. Combining both SNP and CNVs in a machine learning approach gave a better overall prediction compared to SNPs alone (area under the curve (AUC) + 2%). The best case prediction was achieved with respect to the psychiatrically screened control population (AUC 0.86). Loss-of-Function (LOF) variants derived from 335 cases and 165 controls were determined after the annotation of variant calls derived from 30 X WGS (illumina xTen). Variants were then filtered by constraint using probability of loss-of-function intolerance(pLI) extracted from the GnomAD database. These variants were further subjected to pathways analysis. While high pLI variants were not enriched in cases (p = 0.55, OR 1.08), pathways analysis suggested these were enriched in neurodevelopmental disorders and schizophrenia.  In the absence of phenotypic association, effect prediction of genomic variation was further integrated using genomic annotation and allelic expression analysis through RNA sequencing in a smaller subset of samples (56 samples), to identify functional variants impacting on gene expression. This revealed quantitative impact of both rare CNVs and LoF variants on gene expression in individuals. Collectively, our integrative multidimensional approach was able to yield significantly more personalised insight into the systems biology of the disorder and expect this will ultimately account for the entire spectrum of variants across various effect sizes that will significantly impact the way we diagnose and treat schizophrenia