SU97 - GENETICALLY INFORMED SUBTYPING OF SCHIZOPHRENIA

Abstract

Background Schizophrenia is a mental disorder with heterogeneous clinical presentation and genetic predisposition. In DSM IV, schizophrenia was classified into five subtypes based on clinical features. Due to its limited application in clinical practice and the lack of biological underpinning, these subtypes were removed in DSM V. In recent years, many large scale genome-wide association studies have been conducted and numerous genetic factors have been identified for various diseases and traits, including schizophrenia. This provides an opportunity to explore subtyping of complex diseases into genetically homogeneous subgroups so that more effective treatments can be developed. Methods We calculated polygenic scores for the cases of the Molecular Genetics of Schizophrenia (MGS) study (N=2,681) for a variety of diseases and traits, and evaluated their potential as classifiers to subtype schizophrenia using k-means clustering algorithm. We verified the cluster structure in the Sweden Study of Schizophrenia (SSS) study (N=2,895). We further validated the clusters with functional data from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) study (N=741). Results Using k-means algorithm, we found that the polygenic scores of neuroticism and blood creatinine level could reliably classify MGS patients into two clusters (subtypes) and the cluster structures had been verified in the patients of SSCC and CATIE studies. Furthermore, using the rich data of neurocognitive and lab tests in the CATIE study, we found that the cluster memberships were associated neurocognitive functions (Wide Range Achievement Test 3rd edition, Reading score: P=0.0015) and lab tests (creatine phosphokinase test: P=7.2×10-7; HDL: P=0.0003; triglycerides: P=0.0015; white blood cells: P=0.0022; Neutrophil cells: 0.0064). Furthermore, by the end of phase 1 trial, the minor subtype (84 out of 650 subjects) had improved positive symptoms (P=0.0479) comparing to the baseline at enrollment. The negative symptoms showed a trend (P=0.0843). Discussion The results indicate that genetically informed subtyping has functional and biological underpinning. More importantly, our subtyping may be linked to treatment outcomes when measured by the symptom presentation. While our results need to be verified, the implication to clinical practice is clear and significant.