Predicting Treatment Response in Schizophrenia With Magnetic Resonance Imaging and Polygenic Risk Score

• DOI: 10.3389/fgene.2022.848205
• Publication Date: 2022-02-02T05:21:51Z
• Authors: Meng Wang; Ke Hu; Lingzhong Fan; Hao Yan; Peng Li; Tianzi Jiang; Bing Liu
• Keywords: Polygenic risk score; Multifactorial Inheritance

Background: Prior studies have separately demonstrated that magnetic resonance imaging (MRI) and schizophrenia polygenic risk score (PRS) are predictive of antipsychotic medication treatment outcomes in schizophrenia. However, it remains unclear whether MRI combined with PRS can provide superior prognostic performance. Besides, the relative importance these measures predictions is not investigated. Methods: We collected 57 patients schizophrenia, all which had baseline genotype data. All received approximately 6 weeks treatment. Psychotic symptom severity was assessed using Positive Negative Syndrome Scale (PANSS) at follow-up. divided into responders ( N = 20) or non-responders 37) based on their percentages PANSS total reduction were above below 50%. Nine categories PRSs 145 different p -value thresholding ranges calculated. trained machine learning classifiers predictors to identify a patient responder non-responder. Results: The extreme gradient boosting (XGBoost) technique applied build binary classifiers. Using leave-one-out cross-validation scheme, we achieved an accuracy 86% features. Other metrics also estimated, including sensitivity (85%), specificity (86%), F1-score (81%), area under receiver operating characteristic curve (0.86). found excluding single feature category gray matter volume (GMV), amplitude low-frequency fluctuation (ALFF), surface curvature could lead maximum drop 10.5%. These three contributed more than half top 10 important removing features caused modest (8.8%), least decrease (1.8%) among categories. Conclusions: Our classifier both stable biased predicting either Combining measures, certain extra power exhibited medium predictions, lower GMV, ALFF, curvature, but higher cortical thickness, volume, sulcal depth. findings inform contributions