29.3 TOPOLOGICAL ANALYSIS ENABLES PREDICTION OF FUNCTIONAL AND SOCIAL OUTCOMES IN EARLY PSYCHOSIS

Abstract

Abstract Background Predicting functional outcome is a major challenge in service care of early psychosis patients (EPP). We investigated clinical and metabolic features that enable objective assessment of the risk of poor functional outcomes in early psychosis patients (EPP), using state-of-the-art topological data analysis (TDA). The guiding philosophy of TDA is that the shape of large data sets, encoded by a mathematical signature, should reveal important relations among the data points. Our primary TDA tool is Mapper, which provides unsupervised multivariate pattern analysis of high-dimensional data, producing a compressed visual representation of the data giving a strong indication of where to look for meaningful clustering. Application of Mapper has already led to a number of remarkable results, including the identification of a new subtype of breast cancer and of two phenotypically distinct types of fragile X syndrome. Methods Recruitment: 101 EPP with informed consent were recruited from the Lausanne “Treatment and early Intervention in Psychosis Program”. They had reached the psychosis threshold on the CAARMS scale. Patients were assessed at entrance of the program by PANSS and MADRS, functioning by SOFAS and GAF scales. After 3 years of clinical follow-up, EPP were evaluated for symptoms levels, social and occupational functioning. Metabolomics: Plasma extracts at the program begin were analyzed by liquid-chromatography-mass spectrometry. Features data (m/z, retention time, integrated ion intensity) were extracted using xMSanalyser v1.3.5 with apLCMS v5.9.4. Mapper: Using the Mapper algorithm, we built a topological map representing the EP cohort through the lens of the PANSS scores at baseline. The mapper algorithm captures the shape of high dimensional data through a graph. Nodes represent patients or groups of patients. Edges between nodes represent shared patients. Results Three distinct clusters of patients could be identified: Group A, characterized by low levels of negative (N2) and depression symptoms (G6); group B, characterized by high levels of positive symptoms (P1, 6) having the worst outcome; and group C with high levels of negative (N2) and depression symptoms. Comparison of the groups after 3-years of clinical follow-up showed that group A displayed significantly better social and occupational functioning. The same grouping and predictions were confirmed in a second independent cohort of EPP. Inclusion of the metabolomics data revealed that group A had a distinct profile, which may serve as predictor of functional outcome in EEP. Conclusions These novel findings suggest that topological analysis of a combination of symptoms and blood metabolic profile might contribute to the prediction of functional outcome at early stages of psychosis.