O1.7. TRANSLATING TRANSCRIPTOME DATA MINING INTO NEUROBIOLOGICAL AND CLINICAL READOUTS.

Background Pioneering views at the dawn of the GWAS era held that nodes of genetic convergence downstream of any individual molecule per se regulate genetic networks subserving the core neurophysiological elements affected in schizophrenia (SCZ [1]). These nodes of genetic convergence might be, for example, transcription factors or miRNAs representing entry points into biologically valid pathways that affect cellular, systems-level and behavioral phenotypes. Since many of the SCZ risk variants are non-coding and may control gene expression, it is plausible that the regulatory elements of gene expression represent a mechanism of risk. Using Weighted Gene Co-expression Network Analysis (WGCNA), we have previously shown that genetic networks including SCZ risk genes indexed via co-expression quantitative traits loci (co-eQTLs) are associated with core neurophysiological systems-level processes affected in SCZ, like working memory brain activity patterns [2]. However, genetic regulatory elements linking risk genes with altered regulation of gene expression at multiple risk loci are still missing. Methods In two RNA sequencing studies, we first identified a gene co-expression module enriched for miR-137 targets and for SCZ risk, validated it in a mouse cell model, and found that its co-eQTLs are associated with brain activity. In our first study, we computed polygenic scores obtained either combining co-eQTLs associated with target genes of miR-137 or SCZ risk variants within target genes of miR-137; then, we compared their effects on two neurobiological phenotypes of SCZ, i.e. brain activation during working memory and emotion recognition. In our second study, we adopted an approach agnostic of regulators and identified gene co-expression modules enriched for SCZ risk genes. Finally, we associated SCZ risk modules with treatment response to olanzapine in two patients' cohorts. Results We found different neurobiological readouts of the two scores, i.e. polygenic risk for SCZ was associated with working memory, while co-expression scores were associated with emotion recognition. These results suggest that co-eQTLs index a different signal compared to SCZ risk variants, although both are associated with SCZ phenotypes. In the second study, SCZ risk genes were co-expressed to a significant extent and were co-regulated by miR-101, miR-374, and miR-28. Furthermore, in two independent samples of patients with SCZ, the co-eQTLs of a SCZ risk module were associated with the inter-individual variation between patients in terms of positive symptoms decrease following 1-month treatment. Discussion These findings suggest that genetic risk for SCZ translates into illness via gene co-regulation pathways mediated by molecular factors potentially susceptible to genetic influences. Furthermore, these results suggest that data mining of the transcriptome can effectively stratify patients in terms of their neurobiological and clinical phenotypes. This is relevant to prospective approaches entailing personalized pharmacological treatment. References: 1. Harrison, P. J. & Weinberger, D. R. Molecular psychiatry 10, 40–68; (2005). 2. Pergola, G. et al. Translational psychiatry 7, e1006; (2017). [ABSTRACT FROM AUTHOR]