PT  - JOURNAL ARTICLE
AU  - Anna R. Docherty
AU  - Arden Moscati
AU  - Daniel E. Adkins
AU  - Gemma T. Wallace
AU  - Guarav Kumar
AU  - Brien P. Riley
AU  - Aiden Corvin
AU  - F. Anthony O’Neill
AU  - Michael Gill
AU  - Kenneth S. Kendler
AU  - Patrick F. Sullivan
AU  - Ayman H. Fanous
AU  - Silviu-Alin Bacanu
TI  - Proof of concept: Molecular prediction of schizophrenia risk
AID  - 10.1101/129213
DP  - 2017 Jan 01
TA  - bioRxiv
PG  - 129213
4099  - http://biorxiv.org/content/early/2017/04/30/129213.1.short
4100  - http://biorxiv.org/content/early/2017/04/30/129213.1.full
AB  - Question To what extent do global polygenic risk scores (PRS), molecular pathway-specific PRS, complement component (C4) gene expression, MHC loci, sex, and ancestry jointly contribute to risk for schizophrenia-spectrum disorders (SZ)?Findings Global polygenic risk for schizophrenia, sex, and their interaction most robustly predict risk in a classification and regression tree model, with highest risk groups having 50/50 chance of SZ.Meaning Psychometric risk indicators, such as prodromal symptom assessments, may be enhanced by the examination of genetic risk metrics. Preliminary results suggest that of genetic risk metrics, global polygenic information has the most potential to significantly aide in the prediction of SZ.Importance Schizophrenia (SZ) has a complex, heterogeneous symptom presentation with limited established associations between biological markers and illness onset. Many (gene) molecular pathways (MPs) are enriched for SZ signal, but it is still unclear how these MPs, global PRS, major histocompatibility complex (MHC) complement component (C4) gene expression, and MHC loci might jointly contribute to SZ and its clinical presentation. It is also unclear whether sex or ancestry interacts with these metrics to increase risk in certain individuals.Objective To examine multiple genetic metrics, sex, and their interactions as possible predictors of SZ risk. Genetic information could aid in the clinical prediction of risk, but it is still unclear which genetic metrics are most promising, and how sex interacts with genetic risk metrics.Design, Setting, and Participants To examine molecular risk in a proof-of-concept study, we used the Wellcome Trust case-control cohort and classified cases as a function of 1) polygenic risk score (PRS) for both whole genome and for 345 implicated molecular pathways, 2) predicted C4 expression, 3) SZ-relevant MHC loci, 4) sex, and 5) ancestry.Main Outcomes and Measures PRSs, C4 expression, SZ-relevant MHC loci, sex, and ancestry as joint risk factors for SZ.Results Recursive partitioning yielded 15 molecular risk classes and retained as significant psychosis classifiers only sex, genome-wide SZ polygenic risk, and one MP PRS. Sex was the most robust classifier in a stepwise regression, and there was a significant interaction of sex with SZ PRS on case status, suggesting males have a lower polygenic risk threshold. By down-sampling case proportion to 1% and 1.4% population base rates in males and females, respectively, high-risk subtypes defined by this model had roughly a 52% odds of developing SZ (individuals with SZ PRS elevated by 2.6 SDs; incidence = 51.8%).Conclusions and Relevance This proof-of-concept suggests that global SZ PRS, sex, and their interaction are robust predictors of risk and that males have a lower PRS threshold for onset. Implications for the integration of these metrics with psychometrically-identified risk are discussed.