Predicting antipsychotic responsiveness using a machine learning classifier trained on plasma levels of inflammatory markers in schizophrenia

Predicting antipsychotic responsiveness using a machine learning classifier trained on plasma levels of inflammatory markers in schizophrenia | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Predicting antipsychotic responsiveness using a machine learning classifier trained on plasma levels of inflammatory markers in schizophrenia Jimmy Lee, Jie Yin Yee, Ser-Xian Phua, Yuen Mei See, Anand Andiappan, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4604742/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 14 Feb, 2025 Read the published version in Translational Psychiatry → Version 1 posted 12 You are reading this latest preprint version Abstract Background In this research study, we apply machine learning techniques to navigate the multifaceted landscape of schizophrenia. Our method entails the development of predictive models, emphasizing peripheral inflammatory biomarkers, which are classified into treatment response subgroups: antipsychotic-responsive, clozapine-responsive, and clozapine-resistant. Methods The study comprises 146 schizophrenia patients (49 antipsychotics-responsive, 68 clozapine-responsive, 29 clozapine-resistant) and 49 healthy controls. Protein levels of immune biomarkers were quantified using the Olink Target 96 Inflammation Panel. To predict labels, a support vector machine classifier is trained on the Olink data matrix and evaluated via leave-one-out cross-validation. Associated protein biomarkers are identified via recursive feature elimination. Findings We constructed three separate predictive models for binary classification: one to discern healthy controls from individuals with schizophrenia (AUC = 0.74), another to differentiate individuals who were responsive to antipsychotics (AUC = 0.88), and a third to distinguish treatment-resistant individuals (AUC = 0.78). Employing machine learning techniques, we identified features capable of distinguishing between treatment response subgroups. Interpretation In this study, support vector machine demonstrates the power of machine learning to uncover subtle signals often overlooked by traditional statistics. Unlike t-tests, it handles multiple features simultaneously, capturing complex data relationships. Chosen for simplicity, robustness, and reliance on strong feature sets, its integration with artificial intelligence techniques like SHapely Additive exPlanations enhances model interpretability, especially for biomarker screening. This study highlights the potential of integrating machine learning techniques in clinical practice. Not only does it deepen our understanding of schizophrenia's heterogeneity, but it also holds promise for enhancing predictive accuracy, thereby facilitating more targeted and effective interventions in the treatment of this complex mental health disorder. Health sciences/Biomarkers/Predictive markers Health sciences/Diseases/Psychiatric disorders/Schizophrenia Artificial Intelligence (AI) Inflammation Machine Learning (ML) Peripheral Biomarkers Treatment Resistant Schizophrenia (TRS) Figures Figure 1 Figure 2 Figure 3 1. Introduction Schizophrenia is a heterogeneous disorder characterised by its diverse manifestations in presentation, illness trajectory among individuals and treatment outcomes 1 . This heterogeneity challenges efforts to personalize treatment and derive accurate prognosis 1 . The Working Group on Classification of Psychotic Disorders for ICD-11 advocated the removal of traditional subtypes such as paranoid and catatonic as they lack clinical utility in everyday clinical practice 2 . However, these advocated changes are insufficient as classification of schizophrenia still heavily depends on symptom clusters. The introduction of severity and specifiers may pose more hurdles than benefits for busy clinicians. Additionally, the reliability and predictive validity of these additions remain uncertain and may not surpass the previously abandoned subtypes. Therefore, the identification of schizophrenia subtypes continues and involves various strategies such as biological markers, clinical staging and treatment response, each presenting unique challenges 3 . Despite the availability of numerous typical and atypical antipsychotics, except for clozapine's effectiveness in refractory schizophrenia, differences in their efficacy remain inconclusive 4 . Present evidence suggests three distinct antipsychotic response levels: to antipsychotics other than clozapine, to clozapine itself, and a suboptimal response to both. Treatment guidelines typically mandate two unsuccessful antipsychotic trials before fulfilling criteria for treatment resistance and hence clozapine eligibility 2 . However, for those with a suboptimal response to clozapine, there's limited evidence on effective interventions, hinting at a potential "ultraresistant" form of schizophrenia 2 . The significance of immune dysregulation in schizophrenia has garnered research and clinical interest 5 . There is accumulating evidence of immune abnormalities in blood, cerebrospinal fluid, central nervous system, including alterations in immune cell numbers, inflammatory markers, and antibody levels in schizophrenia 6 . This connection presents a compelling avenue for biomarker identification that might serve as vital indicators for clinical use, e.g. in identifying or prognosticating treatment response or resistance 6 . The exploration of immune-related biomarkers thus holds promise for more personalised and effective treatment approaches in schizophrenia. Due to the heterogeneous and complex nature of schizophrenia, subtyping schizophrenia patients before identification of biomarkers is a more reasonable approach 7 . Machine learning (ML) techniques offer several advantages: Firstly, ML algorithms can sieve through high dimensional and heterogenous nature of biological data, which often contain many variables 8 . Through advanced feature selection and dimensionality reduction methods, ML algorithms can identify relevant patterns and associations within the data, even in cases where traditional statistical methods may struggle 8 . Recently, ML algorithms have gained traction in scrutinising schizophrenia patient data e.g., to predict clinical progress, differentiate between schizophrenia patients and healthy individuals, based on neuroimaging data, clinical information, and biological markers 8 – 11 . Building on this momentum, the objective of this study is to employ ML on blood-based measurements of inflammatory markers to train models for categorising individuals into three pharmacological subtypes of schizophrenia: antipsychotic responsive (ARE), clozapine responsive (CRE) and clozapine resistant (CRT). Through these models, we aim to enhance our understanding of schizophrenia subtypes, so to contribute to more personalised and effective treatment strategies. 2. Materials and Methods 2.1. Study settings and subjects The cross-sectional study was conducted at the Institute of Mental Health, Singapore. Individuals living with schizophrenia, aged 21 to 80 were enrolled in the current study. Participants are categorised into three pharmacological subtypes, namely: ARE, CRE and CRT. A set of healthy control (HCL) is also included in this study for comparative analysis. The Clinical Global Impression – Schizophrenia scale (CGI-SCH) assesses positive, negative, depressive, cognitive, and overall symptom severity, and was employed to assess symptom severity in the present study 12 – 14 . In the present study, positive symptom remission is indicated by a CGI-SCH score of 3 and lesser on the positive item, reflecting low levels of illness symptoms with little impairment in patient’s usual social and occupational roles. ARE was defined by the ability to attain CGI-SCH positive symptom remission on non-clozapine antipsychotic agents 12 . Both CRE and CRT refers to individuals with treatment resistant schizophrenia (TRS), characterised by failure of 2 or more unsuccessful antipsychotic trials and subsequent prescription of clozapine. CRE or CRT was defined by the attainment of positive symptom remission after at least 8 weeks of clozapine treatment. Individuals selected as HCL have no past or current history of mental illness or neurological disorder and no first-degree relatives with diagnosed psychiatric disorders. Participants were excluded if they had a history of traumatic brain injury or immune-mediated disorders, had no febrile illness and were not on anti-inflammatory agents for at least one week before study enrolment. Ethics approval for this study was provided by the National Healthcare Group’s Domain Specific Review Board. Written informed consent was obtained from all study participants. 2.2. Data and clinical information Socio-demographic data such as age, sex, ethnicity, and smoking status were collected from all study participants. Anthropometric measurements such as weight and height were collected during the site visit. Four study clinicians ascertained the diagnosis of schizophrenia using the Structured Clinical Interview for DSM-5 (SCID-5) and assessed symptom severity on the CGI-SCH 12 . Each clinician had at least 5 years of clinical experience in psychiatry and attained an inter-rater concordance of > 0.9 on the CGI-SCH. Prescription information for antipsychotics was obtained from participants’ medical records. 2.3. Sample collection and measurement of protein expression of inflammatory markers A sample of venous blood was collected from all study participants into CPT™ blood collection tube. Whole blood was centrifuged at 1700g for 20 minutes for plasma collection which was stored in -80°C. Plasma inflammatory markers were assessed using the commercially available Olink Target 96 Inflammation from Olink (Uppsala, Sweden). Olink's Proximity Extension Assay technology uses antibody pairs conjugated to unique oligonucleotides and is quantified via PCR. When both antibodies of a pair bind the target protein simultaneously, their respective conjugated oligonucleotides are brought into proximity, facilitating hybridization. The oligonucleotide sequence is then extended by DNA polymerase, amplified, and measured by qPCR to determine the sample's initial protein abundance. Raw analyte expression values after PCR underwent multiple rounds of transformation by Olink, including a log2 transformation, and were returned as normalised protein expression values. 2.4. Missing value imputation and batch effect correction on Olink data Plasma samples were put through the Olink platform in three batches to describe relative serum expression for 96 proteins. Signals for any given protein falling below detection limit were treated as missing values and any protein missing for more than half of samples was excluded. 75 proteins met this criterion for inclusion in the analyses. Remaining missing values were imputed based on mean imputation. Batch correction was performed using Olink recommended procedure https://olink.com/application/data-normalization-and-standardization/ . The remaining samples were divided into training set (n = 79), validation set (n = 80), and test set (n = 38) for model training (see next section). 2.5. Model training and SHapely Additive exPlanations (SHAP) analysis Sample labels (HCL, ARE, CRE, CRT) were reorganised into different tiers to perform three sets of learning tasks, each involving a different ML model (Fig. 1A). The first task (Model 1) differentiates HCL from those individuals living with schizophrenia (HCL vs. ARE & CRE & CRT). The second task (Model 2) resolves antipsychotic response from the schizophrenia group, namely antipsychotic responsive and TRS (ARE vs. CRE & CRT). The final task (Model 3) involves differentiating clozapine responders (CRE) from clozapine resistors (CRT). The support vector machine (SVM) was chosen for its simplicity and efficiency 15 , 16 . The feature selection approach, Recursive feature elimination (RFE) is implemented to select important features 17 – 19 . Selected sets of proteins are referred to as feature sets. Model parameters, including C and gamma (penalty terms) were tuned to reduce overfitting to the training data. Other parameters such as the class weights were tuned to mitigate the effect of class imbalance of the dataset. Parameter tuning was optimised via five-fold Grid Search cross-validation (GridSearchCV). The validation set is then deployed to compare and evaluate feature sets and parameters. This step minimises overfitting and increases generalization based on receiver operating characteristics (ROC) score. Once the model is optimised given both the training set and validation set, the test set is used to score the models on ROC scores and balanced accuracy. To analyse our models, SHAP was used to explain models and identify discriminant proteins 20 . To interpret the SVM model using SHAP, contributions of each protein feature from each sample were calculated as SHAP values using KernelExplainer. As model coefficients cannot be directly extracted from a SVM model with a kernel, the feature importance is inferred using these SHAP values. SHAP values can be interpreted as a measure to how much each feature from each sample contributes to the outcome of the prediction (towards a positive or negative prediction) based on the model. Thus, a good performance from the ML model is tantamount to the reliability of the SHAP values generated. All ML tasks were performed using scikit-learn (version 1.1.2) in the Python environment. 2.6. Statistical analysis To evaluate the statistical significance of protein expressions, we use the two-sample t -test. Multiple test correction was performed using the Benjamini-Hochberg procedure, with the family-wise error rate cutoff at 0.05). To access the statistical significance of difference between pharmacological subtypes, Wilcoxon rank-sum test for non-normal distributed data for age and BMI, and Chi-squared test for categorial variables such as sex, smoking status, and ethnicity. 3. Results 3.1. Clinical characteristics The study sample comprised 49 individuals as HCL, a group of people with schizophrenia, categorised into 3 groups − 49 ARE, 68 CRE and 29 CRT. Characteristics of the study sample are summarised in Table 1 . BMI and age differed significantly between the groups. ARE had the highest BMI at 26.9 ( P = 0.001) and CRT was the oldest amongst all (41.6 years, P = 0.015). From the clinical assessments, CRT exhibited the most severe symptoms among the four groups evaluated. We observed no differences in sex and ethnicity in all four groups. A detailed breakdown of demographics for training, testing and validation sets is available in Supplementary S1. Table 1 - Demographic and clinical data of study participants Group HCL (n=49) ARE (n=49) CRE (n=68) CRT (n=29) P 1 BMI 23.8 (4.3) 26.9 (6.8) 26.4 (4.5) 24.0 (4.6) 0.001 Age, years 34.1 (10.9) 39.2 (10.3) 38.4 (10.2) 41.6 (12.7) 0.015 Sex 0.358 Male 27 26 41 21 Female 22 23 27 8 Ethnicity 0.478 Chinese 43 35 57 26 Malay 4 6 4 1 Indian 1 7 6 2 Others 1 1 1 0 CGI-SCH Positive - 1.7 (0.8) 2.2 (0.8) 4.4 (0.6) 0.001 Negative - 1.4 (0.6) 2.1 (1.0) 3.1 (1.0) < 0.001 Depressive - 1.5 (0.7) 1.4 (0.5) 1.8 (0.9) 0.026 Cognitive - 1.4 (0.6) 1.7 (0.7) 2.0 (0.9) < 0.001 Overall - 1.9 (0.7) 2.5 (0.9) 4.3 (0.7) < 0.001 1 Statistical significance of differences was calculated using Kruskal Wallis test or Person's chi-squared test (for sex and ethnicity). Figures in bold and red indicate significant difference at P <0.05 level. Abbreviations: Body Mass Index (BMI); Clinical Global Impression-Schizophrenia scale (CGI-SCH) 3.2. Predictive modelling for schizophrenia status, antipsychotic response and clozapine response In Fig. 2A, the ROC curves illustrate the sensitivities of each classification model. Model 1, which differentiates HCL from SCZ, has a ROC of 0.74, Model 2, distinguishing ARE from TRS, has an ostensibly higher ROC of 0.88. Model 3, which discerns clozapine responsiveness, showed a ROC of 0.78. These results show contrasting outcomes from the ML models compared to standard differential expression analysis using statistical techniques (Fig. 2). The prediction model was able to reasonably differentiate clozapine responsive and clozapine resistant samples, whereas our statistical analysis shows that there is no statistically significant difference between the clozapine responsive and clozapine resistant group in any proteins. Overall, our models can perform reasonably well on their classification tasks. 3.3. Protein Profiles Olink inflammation panel was employed to detect the relative differences in inflammation-related protein expression levels between the study groups described in Fig. 1B. Among the 75 analysed proteins were 44 cytokines and chemokines, 11 cell surface receptors and signaling molecules, 7 enzymes and proteases, 11 growth factors and related molecules, and 2 other regulators and proteins (list of proteins seen in Supplementary S2). All 75 protein markers were used to build predictive models as elaborated in Fig. 2B. 3.4. Comparison of protein levels across groups Figure 2B refers to the protein markers with statistically significant fold changes for comparisons between HCL and cases. 30 proteins exhibited significantly distinct levels between HCL and cases (i.e., ARE & CRE & CRT) (full list of proteins with fold change and P -values in Supplementary S2). Notably, CST5, DNER, SCF, and TWEAK displayed downregulated levels in individuals with schizophrenia, while CCL19, CCL3, CD5, CDCP1, CSF-1, FGF-21, FGF-23, Flt3L, HGF, IL10, IL-10RB, IL-12B, 12-15RA, IL-17C, IL18, IL6, MCP-3, MMP-1, MMP-10, SLAMF1, TGF-alpha, TNF, TNFRSF9, TNFSF14, and VEGFA exhibited upregulated levels. Most of these dysregulated immune biomarkers share functions related to immune regulation, inflammation, and cell signaling. They act on common pathways and networks such as immune cell recruitment and activation, T cell differentiation and proliferation, modulation of cytokine signaling, and cell growth and survival. Figure 2C refers to the protein markers with statistically significant fold change for comparison between antipsychotic responsive and TRS groups. Five proteins (CCL25, CD5, CST5, MMP-10, and TNFRSF9) showed significant differences between antipsychotic responsive and TRS groups (ARE vs CRT + CRE); CCL25 emerged as the exclusive marker unique to antipsychotic responsiveness. No significantly different protein markers were identified between clozapine responsive. 4. Discussion Subgrouping individuals according to their response to antipsychotic treatment has potential for identifying clinically relevant prognostic markers. This could lead to effective personalised treatments which will improve clinical outcomes in individuals with schizophrenia. However, this remains a significant challenge due to the intricate and multifactorial nature of the condition. To the best of our knowledge, this study marks an early effort to develop a predictive model for categorizing antipsychotic responsiveness. 4.1. Model prediction There have been numerous attempts to develop predictive models for precision psychiatry using ML and deep learning algorithms. However, most are trained on neuroimaging, speech patterns and digital features 21 – 23 . Only a handful were based on molecular biomarker data 8 . The present study was able to achieve a reasonable performance of 0.74 to distinguish people with and without schizophrenia, and 0.88 and 0.78 respectively for differentiating responsiveness to antipsychotics and clozapine. However, the drop in performance for model 3 (differentiating clozapine responsiveness) could be due to small sample size. Regardless of accuracy and performance, all models still require external validation or practical implementation in clinical settings. So far, very few prediction tools developed in mental health research have progressed into clinical use, with the prediction of medication response in schizophrenia still in its early stages 24 . Two studies focused on predicting response to medications commonly used in bipolar disorder, namely lithium and quetiapine, utilizing baseline sociodemographic, clinical, and family history information 25 , 26 . Although both studies developed models that performed above chance levels, neither model underwent validation in independent samples. In future research, a crucial strategy will be the integration of biological data for comprehensive analyses. Biological data offers objective measurements of physiological processes or biomarkers, enhancing reliability and reducing variability. Furthermore, the ability to longitudinally monitor changes over time through biological data offers valuable insights into disease progression, treatment efficacy, and individual health trajectories. Therefore, integrating biological data holds significant potential to enhance overall model performance and deepen our understanding of the genetic foundations of mental illnesses. 4.2. Inflammatory markers Using ML, we identified 18 proteins to be the most impactful features to differentiate healthy individuals from people with schizophrenia. These investigated proteins were involved in inflammatory processes, such as Jak-STAT, NF-κB, mitogen-activated protein kinases (MAPK), RAS, and TNF signaling pathway, cytokine-cytokine, and receptor interaction pathways 27 . 11 out of these 18 proteins (CD5, CST5, DNER, IL-10, IL-10RB, IL-12B, IL-15RA, IL-17C, MCP-3, MMP-10, and SCF) exhibited significant difference in protein expression levels between healthy individuals and schizophrenia group. There were reports on differential levels of immune biomarkers during different stages of psychosis (prodromal, first-episode psychosis and chronic schizophrenia against healthy individuals). Results from the present study are concordant with reported findings – higher protein expression levels of commonly studied markers such as CCL23, IL-6, IL-8, IL-10, IL-12, IL-15, TNF and TGF are observed in individuals with schizophrenia when compared to healthy individuals 28 , 29 . As chemokines, growth factors, cell surface receptors traditionally, received less attention than cytokines, reports of such group of inflammatory markers in psychosis or schizophrenia are limited and often inconsistent. Nonetheless, we observed a general upward trend for similar markers for individuals with schizophrenia in the present study 30 , 31 . Several meta-analyses on associations between immunological markers and antipsychotics response were limited to pre- and post-treatment in first episode psychosis 32 , 33 . Among numerous immune biomarkers studied, IL-1β, IL-6, and TGF-β were consistently reported to have significant reductions post-antipsychotic treatment. Thus far, only one meta-analysis conducted to evaluate the immune alterations induced by clozapine which reported a down-regulation of IL-6 levels in patients who received clozapine, compared to other antipsychotics 34 . Although the present study had similar down-regulation of IL-6 (log2 fold change of -0.112) in TRS compared to ARE, this fold change did not meet statistical significance. Out of the 8 immune markers identified via ML, only IL-18 and TNF from this study was associated with response to amisulpride, risperidone, olanzapine, and/or quetiapine 33 . The remaining six markers, CCL25, CST5, MCP-2, MCP-4, PD-L1, and TNFRSF9 require further investigation. 4.3. The value of ML for classifying pharmacological subtypes ML methods such as the SVM used in this study can discern subtle signals that may not be identified through traditional statistical testing. Unlike conventional statistical methods like the t-tests, ML methods can model multiple features simultaneously and capture relationships between features, which might be overlooked. The SVM was selected for its simplicity, robustness, and its reliance on a good feature set. When coupled with the explainable AI approach SHAP, we were able to improve model interpretability. Based on Fig. 3, several statistically significant features had less predictive power compared to features that were not statistically significant based on the mean absolute SHAP value. Hence, this approach can be powerful for biomarker screening. Contemporary methods like ensemble methods (such as bagging and boosting) are recognised for their superior prediction performance score and popularity. However, in this context, they may not be appropriate due to our limited sample and feature size. 4.4. Strengths and limitations Strengths. The sample population is from a single site in a tertiary healthcare institute, which enabled enrollment of a well-characterised clinical sample and reduce heterogeneity in data and sample collection. Adopting plasma proteins enables high throughout analysis, allowing for the simultaneous identification and measurement of a vast array of proteins. Limitations. The inflammation panel in the current study is a general list and not tailored for psychosis or mental disorders. We were unable to enroll a drug-naïve group to investigate the impact of medication exposure on peripheral immune profiles. Our study lacked longitudinal evaluation, essential for understanding the consistency of immune profiles among pharmacological subtypes, as well as the influence of factors such as exposure to infectious agents or alterations in medications or treatment regimes. We demonstrated the efficacy of using ML to predict pharmacological subtypes of schizophrenia based on blood-based inflammatory biomarkers. Our study shows that extracted biomarkers from ML models exhibit different expression levels but do not necessarily attain statistical significance possibly due to limited sample sizes and clinical heterogeneity. Despite these limitations, the functional characterization of ML selected biomarkers is supported by contemporary literature, which validates their associations with biological pathways or processes that are relevant to schizophrenia. Hence, ML-based models can be useful for predicting treatment outcomes and yield meaningful markers for understanding underlying mechanisms. Declarations Funding This research is supported by the Singapore Ministry of Health’s National Medical Research Council under its MOH-CSAINV17nov-0004. Credit authorship contribution statement JL designed the study and obtained the funding. JYY and YMS managed the study, supported recruitment and data collection. JYY and SXP analysed the data. AKA assisted with Olink data generation. JYY, SXP, WWBG and JL interpreted the results. JYY and SXP drafted the manuscript. WWBG and JL revised the manuscript. All authors have read and approved the final version of the manuscript. Declarations of Competing Interest JL has received honoraria from Sumitomo Pharmaceuticals, Lundbeck Singapore, Otsuka Pharmaceutical and Janssen Pharmaceutical. All other authors declare that they have no conflict of interest. Acknowledgement The authors would like to express their sincere gratitude to all the participants who participated in this study. References Patel, K. R., Cherian, J., Gohil, K. & Atkinson, D. Schizophrenia: overview and treatment options. P t 39 , 638-645 (2014). Farooq, S., Agid, O., Foussias, G. & Remington, G. Using treatment response to subtype schizophrenia: proposal for a new paradigm in classification. Schizophr Bull 39 , 1169-1172, doi:10.1093/schbul/sbt137 (2013). Zhu, L. et al. The machine learning algorithm for the diagnosis of schizophrenia on the basis of gene expression in peripheral blood. Neuroscience letters 745 , 135596, doi:10.1016/j.neulet.2020.135596 (2021). Nucifora, F. C., Jr., Woznica, E., Lee, B. J., Cascella, N. & Sawa, A. Treatment resistant schizophrenia: Clinical, biological, and therapeutic perspectives. Neurobiology of disease 131 , 104257, doi:10.1016/j.nbd.2018.08.016 (2019). Schwarz, M. J., Müller, N., Riedel, M. & Ackenheil, M. The Th2-hypothesis of schizophrenia: a strategy to identify a subgroup of schizophrenia caused by immune mechanisms. Med Hypotheses 56 , 483-486, doi:10.1054/mehy.2000.1203 (2001). Miller, B. J. & Goldsmith, D. R. Evaluating the Hypothesis That Schizophrenia Is an Inflammatory Disorder. Focus (American Psychiatric Publishing) 18 , 391-401, doi:10.1176/appi.focus.20200015 (2020). Cortes-Briones, J. A., Tapia-Rivas, N. I., D'Souza, D. C. & Estevez, P. A. Going deep into schizophrenia with artificial intelligence. Schizophrenia research 245 , 122-140, doi:10.1016/j.schres.2021.05.018 (2022). Kozyrev, E. A. et al. Building Predictive Models for Schizophrenia Diagnosis with Peripheral Inflammatory Biomarkers. Biomedicines 11 , doi:10.3390/biomedicines11071990 (2023). Chilla, G. S., Yeow, L. Y., Chew, Q. H., Sim, K. & Prakash, K. N. B. Machine learning classification of schizophrenia patients and healthy controls using diverse neuroanatomical markers and Ensemble methods. Scientific reports 12 , 2755, doi:10.1038/s41598-022-06651-4 (2022). Shim, M., Hwang, H. J., Kim, D. W., Lee, S. H. & Im, C. H. Machine-learning-based diagnosis of schizophrenia using combined sensor-level and source-level EEG features. Schizophrenia research 176 , 314-319, doi:10.1016/j.schres.2016.05.007 (2016). Ferrara, M. et al. Machine Learning and Non-Affective Psychosis: Identification, Differential Diagnosis, and Treatment. Current psychiatry reports 24 , 925-936, doi:10.1007/s11920-022-01399-0 (2022). Haro, J. M. et al. The Clinical Global Impression-Schizophrenia scale: a simple instrument to measure the diversity of symptoms present in schizophrenia. Acta psychiatrica Scandinavica. Supplementum , 16-23, doi:10.1034/j.1600-0447.107.s416.5.x (2003). Lee, J. et al. Subtyping Schizophrenia by Treatment Response: Antipsychotic Development and the Central Role of Positive Symptoms. Canadian journal of psychiatry. Revue canadienne de psychiatrie 60 , 515-522, doi:10.1177/070674371506001107 (2015). Lambert, M., Karow, A., Leucht, S., Schimmelmann, B. G. & Naber, D. Remission in schizophrenia: validity, frequency, predictors, and patients' perspective 5 years later. Dialogues in clinical neuroscience 12 , 393-407, doi:10.31887/DCNS.2010.12.3/mlambert (2010). Siemers, F. M. & Bajorath, J. Differences in learning characteristics between support vector machine and random forest models for compound classification revealed by Shapley value analysis. Scientific reports 13 , 5983, doi:10.1038/s41598-023-33215-x (2023). Scott Lundberg, S.-I. L. A Unified Approach to Interpreting Model Predictions. arxiv (2017). Guyon, I., Weston, J., Barnhill, S. & Vapnik, V. Gene Selection for Cancer Classification using Support Vector Machines. Machine Learning 46 , 389-422, doi:10.1023/A:1012487302797 (2002). Li, Z., Xie, W. & Liu, T. Efficient feature selection and classification for microarray data. PloS one 13 , e0202167, doi:10.1371/journal.pone.0202167 (2018). Wang, Z., Meng, Z. & Chen, C. Screening of potential biomarkers in peripheral blood of patients with depression based on weighted gene co-expression network analysis and machine learning algorithms. Frontiers in psychiatry 13 , 1009911, doi:10.3389/fpsyt.2022.1009911 (2022). Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. Advances in neural information processing systems 30 (2017). Rezaii, N., Walker, E. & Wolff, P. A machine learning approach to predicting psychosis using semantic density and latent content analysis. npj Schizophrenia 5 , 9, doi:10.1038/s41537-019-0077-9 (2019). Nenning, K.-H. & Langs, G. Machine learning in neuroimaging: from research to clinical practice. Die Radiologie 62 , 1-10, doi:10.1007/s00117-022-01051-1 (2022). Benoit, J., Onyeaka, H., Keshavan, M. & Torous, J. Systematic Review of Digital Phenotyping and Machine Learning in Psychosis Spectrum Illnesses. Harvard Review of Psychiatry 28 , 296-304, doi:10.1097/hrp.0000000000000268 (2020). Chekroud, A. M. et al. The promise of machine learning in predicting treatment outcomes in psychiatry. World psychiatry : official journal of the World Psychiatric Association (WPA) 20 , 154-170, doi:10.1002/wps.20882 (2021). Nunes, A. et al. Prediction of lithium response using clinical data. Acta psychiatrica Scandinavica 141 , 131-141, doi:10.1111/acps.13122 (2020). Kim, T. T. et al. Predictive modeling for response to lithium and quetiapine in bipolar disorder. Bipolar disorders 21 , 428-436, doi:10.1111/bdi.12752 (2019). Schnorr, I. et al. Inflammatory biotype of ADHD is linked to chronic stress: a data-driven analysis of the inflammatory proteome. Translational psychiatry 14 , 37, doi:10.1038/s41398-023-02729-3 (2024). Momtazmanesh, S., Zare-Shahabadi, A. & Rezaei, N. Cytokine Alterations in Schizophrenia: An Updated Review. Frontiers in psychiatry 10 , 892, doi:10.3389/fpsyt.2019.00892 (2019). Korhonen, L. et al. Multivariate analyses of immune markers reveal increases in plasma EN-RAGE in first-episode psychosis patients. Translational psychiatry 13 , 326, doi:10.1038/s41398-023-02627-8 (2023). Ermakov, E. A., Mednova, I. A., Boiko, A. S., Buneva, V. N. & Ivanova, S. A. Chemokine Dysregulation and Neuroinflammation in Schizophrenia: A Systematic Review. International journal of molecular sciences 24 , doi:10.3390/ijms24032215 (2023). Li, X., Hu, S. & Liu, P. Vascular-related biomarkers in psychosis: a systematic review and meta-analysis. Frontiers in psychiatry 14 , 1241422, doi:10.3389/fpsyt.2023.1241422 (2023). Marcinowicz, P. et al. A Meta-Analysis of the Influence of Antipsychotics on Cytokines Levels in First Episode Psychosis. Journal of clinical medicine 10 , doi:10.3390/jcm10112488 (2021). Orbe, E. B. & Benros, M. E. Immunological Biomarkers as Predictors of Treatment Response in Psychotic Disorders. Journal of personalized medicine 13 , doi:10.3390/jpm13091382 (2023). Martins, P. L. B. et al. Immunoinflammatory and oxidative alterations in subjects with schizophrenia under clozapine: A meta-analysis. European neuropsychopharmacology : the journal of the European College of Neuropsychopharmacology 73 , 82-95, doi:10.1016/j.euroneuro.2023.04.003 (2023). Additional Declarations Jimmy Lee has received honoraria from Sumitomo Pharmaceuticals, Lundbeck Singapore, Otsuka Pharmaceutical and Janssen Pharmaceutical. All other authors declare that they have no conflict of interest. Supplementary Files Supplementarytables.docx Cite Share Download PDF Status: Published Journal Publication published 14 Feb, 2025 Read the published version in Translational Psychiatry → Version 1 posted Editorial decision: revise 19 Aug, 2024 Review # 3 received at journal 30 Jul, 2024 Review # 2 received at journal 22 Jul, 2024 Review # 1 received at journal 22 Jul, 2024 Reviewer # 3 agreed at journal 15 Jul, 2024 Reviewer # 2 agreed at journal 10 Jul, 2024 Reviewer # 1 agreed at journal 09 Jul, 2024 Reviewers invited by journal 09 Jul, 2024 Submission checks completed at journal 24 Jun, 2024 First submitted to journal 24 Jun, 2024 Unknown event 20 Jun, 2024 Editor assigned by journal 19 Jun, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4604742","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":324902310,"identity":"05560236-5b7e-439f-867a-64f38f8141a3","order_by":0,"name":"Jimmy Lee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYBACxmYwZcHAwN7YwEyKFgkGBp6DRGqBAqAWiQQG4rQwt3OnbuZhkJDjl3zcurmAwU5Ot4H98Qv8DuPddhuoxVhydmLb7RkMycZmB3jMLIjRkrjhNlALD8OBxG0HeNgMiNFSv//mQZgW9mdEaUkwkGCEaWEwfkBIy805DBKGM86AHGYA9MthHjN8OhgM+89uu/GGwUaev/34s9s8FXZyZsfbH3/Aq6WBgYGJ9x+MC/IEMwObBD4t8iDH/UATZMZryygYBaNgFIw4AADfDEZ4JB5oDAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-7724-7445","institution":"Institute of Mental Health","correspondingAuthor":true,"prefix":"","firstName":"Jimmy","middleName":"","lastName":"Lee","suffix":""},{"id":324902311,"identity":"0ff8639d-40e8-4c74-bb9e-92e28fa5cc17","order_by":1,"name":"Jie Yin Yee","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"Yin","lastName":"Yee","suffix":""},{"id":324902312,"identity":"978c6e43-f11a-4195-aff4-5014012ca1d2","order_by":2,"name":"Ser-Xian Phua","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ser-Xian","middleName":"","lastName":"Phua","suffix":""},{"id":324902313,"identity":"76fb052e-7bcd-45ce-aee4-3e185df16fe4","order_by":3,"name":"Yuen Mei See","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Yuen","middleName":"Mei","lastName":"See","suffix":""},{"id":324902314,"identity":"c5303860-d8fd-44aa-a749-d906ef1114b7","order_by":4,"name":"Anand Andiappan","email":"","orcid":"https://orcid.org/0000-0002-8442-1544","institution":"Singapore Immunology Network (SIgN)","correspondingAuthor":false,"prefix":"","firstName":"Anand","middleName":"","lastName":"Andiappan","suffix":""},{"id":324902315,"identity":"fed2ce3d-2260-4772-a7c8-6d0016ab204b","order_by":5,"name":"Wilson Goh","email":"","orcid":"","institution":"Nanyang Technological University","correspondingAuthor":false,"prefix":"","firstName":"Wilson","middleName":"","lastName":"Goh","suffix":""}],"badges":[],"createdAt":"2024-06-19 09:20:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4604742/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4604742/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41398-025-03264-z","type":"published","date":"2025-02-14T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":62184393,"identity":"4cb4bba3-1035-4d2d-a8c8-6a70c994dcbc","added_by":"auto","created_at":"2024-08-10 11:41:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":262420,"visible":true,"origin":"","legend":"\u003cp\u003eExperimental approach and experimental design. (A) Hierarchical approach used for model development to determine schizophrenia status, antipsychotic response, and clozapine response. HCL refers to healthy control; ARE refers to antipsychotic responsive; CRE refers to clozapine responsive; CRT refers to clozapine resistant. (B) Validation pipeline using train-test split.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4604742/v1/bd3eca148d9f8f845fc4518e.png"},{"id":62183138,"identity":"d420d669-01a3-46bf-aba2-1b17c4dabdef","added_by":"auto","created_at":"2024-08-10 11:33:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":160940,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Model performance for each classification task to determine pharmacological status. ROC curves for each model - (left) Status classification (HCL v. people with schizophrenia), (middle) Antipsychotic response classification (antipsychotic responsive v. antipsychotics resistant), and (right) Clozapine response classification (clozapine responsive v. clozapine resistant). \u003cstrong\u003e(B)\u003c/strong\u003e Bar graph showing the statistically significant fold change in protein expression of individuals with schizophrenia, compared to HCL. \u003cstrong\u003e(C)\u003c/strong\u003e Bar graph showing the statistically significant fold change in protein expression of individuals on clozapine, compared to individuals on antipsychotics except clozapine.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4604742/v1/d6bf20b4b474f33b34aa3837.png"},{"id":62183141,"identity":"4a7d3296-5552-462f-adba-c219c7ab8512","added_by":"auto","created_at":"2024-08-10 11:33:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":144583,"visible":true,"origin":"","legend":"\u003cp\u003eBeeswarm plot of SHAP-calculation for protein markers identified via ML for predictive models to differentiate treatment response in individuals with schizophrenia. Each datapoint represents a prediction from the test set, the x-axis represents selected proteins, and the y-axis represents the respective impact on the prediction outcome by the model (e.g. towards treatment resistance or treatment positive). Since the datapoints represent the contribution of each sample by each feature to its prediction, they are color coded to its relative protein expression value in the test set. Grey hue along the x-axis denotes proteins that are deemed to be statistically significant. There are no statistically significant features detected between clozapine responsive and clozapine resistant samples. \u003cstrong\u003e(top)\u003c/strong\u003e HCL from participants with schizophrenia, \u003cstrong\u003e(middle) \u003c/strong\u003eAntipsychotic responsiveness (ARE vs TRS), and \u003cstrong\u003e(bottom)\u003c/strong\u003e Individuals with TRS (CRE vs. CRT). Each datapoint represents a sample of each feature, which represents its SHAP value and how it contributes to the final prediction outcome (refer to Supplementary S3 for a comprehensive list of fold changes, P and SHAP values).\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4604742/v1/e16d62c0f29c84d84e356743.png"},{"id":76345258,"identity":"d8a8df8c-c6c6-4703-93b9-7974435ed152","added_by":"auto","created_at":"2025-02-15 08:05:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1474669,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4604742/v1/9428c6dd-88b2-4008-9692-824ee01372a5.pdf"},{"id":62183137,"identity":"0116bce8-7f98-4494-8be1-18f53e397d21","added_by":"auto","created_at":"2024-08-10 11:33:28","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":44206,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytables.docx","url":"https://assets-eu.researchsquare.com/files/rs-4604742/v1/6becb8eaa370879ce5213e62.docx"}],"financialInterests":"\nJimmy Lee has received honoraria from Sumitomo Pharmaceuticals, Lundbeck Singapore, Otsuka Pharmaceutical and Janssen Pharmaceutical. All other authors declare that they have no conflict of interest.","formattedTitle":"Predicting antipsychotic responsiveness using a machine learning classifier trained on plasma levels of inflammatory markers in schizophrenia","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eSchizophrenia is a heterogeneous disorder characterised by its diverse manifestations in presentation, illness trajectory among individuals and treatment outcomes \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. This heterogeneity challenges efforts to personalize treatment and derive accurate prognosis \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. The Working Group on Classification of Psychotic Disorders for ICD-11 advocated the removal of traditional subtypes such as paranoid and catatonic as they lack clinical utility in everyday clinical practice \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. However, these advocated changes are insufficient as classification of schizophrenia still heavily depends on symptom clusters. The introduction of severity and specifiers may pose more hurdles than benefits for busy clinicians. Additionally, the reliability and predictive validity of these additions remain uncertain and may not surpass the previously abandoned subtypes. Therefore, the identification of schizophrenia subtypes continues and involves various strategies such as biological markers, clinical staging and treatment response, each presenting unique challenges \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite the availability of numerous typical and atypical antipsychotics, except for clozapine's effectiveness in refractory schizophrenia, differences in their efficacy remain inconclusive \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Present evidence suggests three distinct antipsychotic response levels: to antipsychotics other than clozapine, to clozapine itself, and a suboptimal response to both. Treatment guidelines typically mandate two unsuccessful antipsychotic trials before fulfilling criteria for treatment resistance and hence clozapine eligibility \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. However, for those with a suboptimal response to clozapine, there's limited evidence on effective interventions, hinting at a potential \"ultraresistant\" form of schizophrenia \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe significance of immune dysregulation in schizophrenia has garnered research and clinical interest \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. There is accumulating evidence of immune abnormalities in blood, cerebrospinal fluid, central nervous system, including alterations in immune cell numbers, inflammatory markers, and antibody levels in schizophrenia \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. This connection presents a compelling avenue for biomarker identification that might serve as vital indicators for clinical use, e.g. in identifying or prognosticating treatment response or resistance \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. The exploration of immune-related biomarkers thus holds promise for more personalised and effective treatment approaches in schizophrenia.\u003c/p\u003e \u003cp\u003eDue to the heterogeneous and complex nature of schizophrenia, subtyping schizophrenia patients before identification of biomarkers is a more reasonable approach \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Machine learning (ML) techniques offer several advantages: Firstly, ML algorithms can sieve through high dimensional and heterogenous nature of biological data, which often contain many variables \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Through advanced feature selection and dimensionality reduction methods, ML algorithms can identify relevant patterns and associations within the data, even in cases where traditional statistical methods may struggle \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Recently, ML algorithms have gained traction in scrutinising schizophrenia patient data e.g., to predict clinical progress, differentiate between schizophrenia patients and healthy individuals, based on neuroimaging data, clinical information, and biological markers \u003csup\u003e\u003cspan additionalcitationids=\"CR9 CR10\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBuilding on this momentum, the objective of this study is to employ ML on blood-based measurements of inflammatory markers to train models for categorising individuals into three pharmacological subtypes of schizophrenia: antipsychotic responsive (ARE), clozapine responsive (CRE) and clozapine resistant (CRT). Through these models, we aim to enhance our understanding of schizophrenia subtypes, so to contribute to more personalised and effective treatment strategies.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Study settings and subjects\u003c/h2\u003e \u003cp\u003eThe cross-sectional study was conducted at the Institute of Mental Health, Singapore. Individuals living with schizophrenia, aged 21 to 80 were enrolled in the current study. Participants are categorised into three pharmacological subtypes, namely: ARE, CRE and CRT. A set of healthy control (HCL) is also included in this study for comparative analysis.\u003c/p\u003e \u003cp\u003eThe Clinical Global Impression \u0026ndash; Schizophrenia scale (CGI-SCH) assesses positive, negative, depressive, cognitive, and overall symptom severity, and was employed to assess symptom severity in the present study \u003csup\u003e\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. In the present study, positive symptom remission is indicated by a CGI-SCH score of 3 and lesser on the positive item, reflecting low levels of illness symptoms with little impairment in patient\u0026rsquo;s usual social and occupational roles. ARE was defined by the ability to attain CGI-SCH positive symptom remission on non-clozapine antipsychotic agents \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Both CRE and CRT refers to individuals with treatment resistant schizophrenia (TRS), characterised by failure of 2 or more unsuccessful antipsychotic trials and subsequent prescription of clozapine. CRE or CRT was defined by the attainment of positive symptom remission after at least 8 weeks of clozapine treatment. Individuals selected as HCL have no past or current history of mental illness or neurological disorder and no first-degree relatives with diagnosed psychiatric disorders. Participants were excluded if they had a history of traumatic brain injury or immune-mediated disorders, had no febrile illness and were not on anti-inflammatory agents for at least one week before study enrolment.\u003c/p\u003e \u003cp\u003eEthics approval for this study was provided by the National Healthcare Group\u0026rsquo;s Domain Specific Review Board. Written informed consent was obtained from all study participants.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Data and clinical information\u003c/h2\u003e \u003cp\u003eSocio-demographic data such as age, sex, ethnicity, and smoking status were collected from all study participants. Anthropometric measurements such as weight and height were collected during the site visit. Four study clinicians ascertained the diagnosis of schizophrenia using the Structured Clinical Interview for DSM-5 (SCID-5) and assessed symptom severity on the CGI-SCH \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Each clinician had at least 5 years of clinical experience in psychiatry and attained an inter-rater concordance of \u0026gt;\u0026thinsp;0.9 on the CGI-SCH. Prescription information for antipsychotics was obtained from participants\u0026rsquo; medical records.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Sample collection and measurement of protein expression of inflammatory markers\u003c/h2\u003e \u003cp\u003eA sample of venous blood was collected from all study participants into CPT\u0026trade; blood collection tube. Whole blood was centrifuged at 1700g for 20 minutes for plasma collection which was stored in -80\u0026deg;C. Plasma inflammatory markers were assessed using the commercially available Olink Target 96 Inflammation from Olink (Uppsala, Sweden). Olink's Proximity Extension Assay technology uses antibody pairs conjugated to unique oligonucleotides and is quantified via PCR. When both antibodies of a pair bind the target protein simultaneously, their respective conjugated oligonucleotides are brought into proximity, facilitating hybridization. The oligonucleotide sequence is then extended by DNA polymerase, amplified, and measured by qPCR to determine the sample's initial protein abundance. Raw analyte expression values after PCR underwent multiple rounds of transformation by Olink, including a log2 transformation, and were returned as normalised protein expression values.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Missing value imputation and batch effect correction on Olink data\u003c/h2\u003e \u003cp\u003ePlasma samples were put through the Olink platform in three batches to describe relative serum expression for 96 proteins. Signals for any given protein falling below detection limit were treated as missing values and any protein missing for more than half of samples was excluded. 75 proteins met this criterion for inclusion in the analyses. Remaining missing values were imputed based on mean imputation.\u003c/p\u003e \u003cp\u003eBatch correction was performed using Olink recommended procedure \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://olink.com/application/data-normalization-and-standardization/\u003c/span\u003e\u003cspan address=\"https://olink.com/application/data-normalization-and-standardization/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe remaining samples were divided into training set (n\u0026thinsp;=\u0026thinsp;79), validation set (n\u0026thinsp;=\u0026thinsp;80), and test set (n\u0026thinsp;=\u0026thinsp;38) for model training (see next section).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Model training and SHapely Additive exPlanations (SHAP) analysis\u003c/h2\u003e \u003cp\u003eSample labels (HCL, ARE, CRE, CRT) were reorganised into different tiers to perform three sets of learning tasks, each involving a different ML model (Fig.\u0026nbsp;1A). The first task (Model 1) differentiates HCL from those individuals living with schizophrenia (HCL vs. ARE \u0026amp; CRE \u0026amp; CRT). The second task (Model 2) resolves antipsychotic response from the schizophrenia group, namely antipsychotic responsive and TRS (ARE vs. CRE \u0026amp; CRT). The final task (Model 3) involves differentiating clozapine responders (CRE) from clozapine resistors (CRT).\u003c/p\u003e \u003cp\u003eThe support vector machine (SVM) was chosen for its simplicity and efficiency \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. The feature selection approach, Recursive feature elimination (RFE) is implemented to select important features \u003csup\u003e\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Selected sets of proteins are referred to as feature sets. Model parameters, including C and gamma (penalty terms) were tuned to reduce overfitting to the training data. Other parameters such as the class weights were tuned to mitigate the effect of class imbalance of the dataset. Parameter tuning was optimised via five-fold Grid Search cross-validation (GridSearchCV).\u003c/p\u003e \u003cp\u003eThe validation set is then deployed to compare and evaluate feature sets and parameters. This step minimises overfitting and increases generalization based on receiver operating characteristics (ROC) score. Once the model is optimised given both the training set and validation set, the test set is used to score the models on ROC scores and balanced accuracy.\u003c/p\u003e \u003cp\u003eTo analyse our models, SHAP was used to explain models and identify discriminant proteins \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. To interpret the SVM model using SHAP, contributions of each protein feature from each sample were calculated as SHAP values using KernelExplainer. As model coefficients cannot be directly extracted from a SVM model with a kernel, the feature importance is inferred using these SHAP values. SHAP values can be interpreted as a measure to how much each feature from each sample contributes to the outcome of the prediction (towards a positive or negative prediction) based on the model. Thus, a good performance from the ML model is tantamount to the reliability of the SHAP values generated.\u003c/p\u003e \u003cp\u003eAll ML tasks were performed using scikit-learn (version 1.1.2) in the Python environment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Statistical analysis\u003c/h2\u003e \u003cp\u003eTo evaluate the statistical significance of protein expressions, we use the two-sample \u003cem\u003et\u003c/em\u003e-test. Multiple test correction was performed using the Benjamini-Hochberg procedure, with the family-wise error rate cutoff at 0.05).\u003c/p\u003e \u003cp\u003eTo access the statistical significance of difference between pharmacological subtypes, Wilcoxon rank-sum test for non-normal distributed data for age and BMI, and Chi-squared test for categorial variables such as sex, smoking status, and ethnicity.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Clinical characteristics\u003c/h2\u003e \u003cp\u003eThe study sample comprised 49 individuals as HCL, a group of people with schizophrenia, categorised into 3 groups \u0026minus;\u0026thinsp;49 ARE, 68 CRE and 29 CRT. Characteristics of the study sample are summarised in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. BMI and age differed significantly between the groups. ARE had the highest BMI at 26.9 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001) and CRT was the oldest amongst all (41.6 years, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.015). From the clinical assessments, CRT exhibited the most severe symptoms among the four groups evaluated. We observed no differences in sex and ethnicity in all four groups. A detailed breakdown of demographics for training, testing and validation sets is available in Supplementary S1.\u003c/p\u003e \n\u003cp style='margin-top:0in;margin-right:0in;margin-bottom:10.0pt;margin-left:0in;font-size:12px;font-family:\"Calibri\",sans-serif;color:#44546A;font-style:italic;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:windowtext;font-style:normal;'\u003eTable\u0026nbsp;\u003c/span\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:windowtext;font-style:normal;'\u003e1\u003c/span\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:windowtext;font-style:normal;'\u003e\u0026nbsp;-\u003c/span\u003e\u003c/strong\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:windowtext;font-style:normal;'\u003e\u0026nbsp;Demographic and clinical data of study participants\u003c/span\u003e\u003c/p\u003e\n\u003ctable style=\"width:460.45pt;border-collapse:collapse;border:none;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eGroup\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border:solid windowtext 1.0pt;border-left: none;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eHCL\u003cbr\u003e\u0026nbsp;(n=49)\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border:solid windowtext 1.0pt;border-left: none;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eARE\u003cbr\u003e\u0026nbsp;(n=49)\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border:solid windowtext 1.0pt;border-left: none;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eCRE\u003cbr\u003e\u0026nbsp;(n=68)\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border:solid windowtext 1.0pt;border-left: none;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eCRT\u003cbr\u003e\u0026nbsp;(n=29)\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border:solid windowtext 1.0pt;border-left: none;padding:0in 5.4pt 0in 5.4pt;height:24.0pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cem\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eP\u003c/span\u003e\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003csup\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eBMI\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e23.8 (4.3)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e26.9 (6.8)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e26.4 (4.5)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e24.0 (4.6)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e0.001\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eAge, years\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e34.1 (10.9)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e39.2 (10.3)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e38.4 (10.2)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e41.6 (12.7)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e0.015\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cu\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eSex\u003c/span\u003e\u003c/u\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;'\u003e0.358\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eMale\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e27\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e26\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e41\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e21\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eFemale\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e22\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e23\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e27\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e8\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cu\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eEthnicity\u003c/span\u003e\u003c/u\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;'\u003e0.478\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eChinese\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e43\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e35\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e57\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e26\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eMalay\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e4\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e6\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e4\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eIndian\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e7\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e6\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e2\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eOthers\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e0\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eCGI-SCH\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003ePositive\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e-\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.7 (0.8)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e2.2 (0.8)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e4.4 (0.6)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e0.001\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eNegative\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e-\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.4 (0.6)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e2.1 (1.0)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e3.1 (1.0)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e\u0026lt; 0.001\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eDepressive\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e-\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.5 (0.7)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.4 (0.5)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.8 (0.9)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e0.026\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eCognitive\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e-\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.4 (0.6)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.7 (0.7)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e2.0 (0.9)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e\u0026lt; 0.001\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:89.7pt;border:solid windowtext 1.0pt;border-top: none;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:right;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eOverall\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:64.9pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e-\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e1.9 (0.7)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:63.8pt;border-top:none;border-left:none;border-bottom: solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e2.5 (0.9)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:71.95pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003e4.3 (0.7)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:106.3pt;border-top:none;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;text-align:center;line-height:normal;'\u003e\u003cstrong\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:red;'\u003e\u0026lt; 0.001\u003c/span\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" style=\"width:460.45pt;border:solid windowtext 1.0pt;border-top:none;padding:0in 5.4pt 0in 5.4pt;height:43.5pt;\"\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003csup\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003e1\u0026nbsp;\u003c/span\u003e\u003c/sup\u003e\u003cspan style='font-size:13px;font-family: \"Arial\",sans-serif;color:black;'\u003eStatistical significance of differences was calculated using Kruskal Wallis test or Person\u0026apos;s chi-squared test (for sex and ethnicity). Figures in bold and red indicate significant difference at \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05 level.\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin-top:0in;margin-right:0in;margin-bottom:0in;margin-left:0in;font-size:11.0pt;font-family:\"Calibri\",sans-serif;line-height:normal;'\u003e\u003cspan style='font-size:13px;font-family:\"Arial\",sans-serif;color:black;'\u003eAbbreviations: Body Mass Index (BMI); Clinical Global Impression-Schizophrenia scale (CGI-SCH)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Predictive modelling for schizophrenia status, antipsychotic response and clozapine response\u003c/h2\u003e \u003cp\u003eIn Fig.\u0026nbsp;2A, the ROC curves illustrate the sensitivities of each classification model. Model 1, which differentiates HCL from SCZ, has a ROC of 0.74, Model 2, distinguishing ARE from TRS, has an ostensibly higher ROC of 0.88. Model 3, which discerns clozapine responsiveness, showed a ROC of 0.78. These results show contrasting outcomes from the ML models compared to standard differential expression analysis using statistical techniques (Fig.\u0026nbsp;2). The prediction model was able to reasonably differentiate clozapine responsive and clozapine resistant samples, whereas our statistical analysis shows that there is no statistically significant difference between the clozapine responsive and clozapine resistant group in any proteins. Overall, our models can perform reasonably well on their classification tasks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Protein Profiles\u003c/h2\u003e \u003cp\u003eOlink inflammation panel was employed to detect the relative differences in inflammation-related protein expression levels between the study groups described in Fig.\u0026nbsp;1B. Among the 75 analysed proteins were 44 cytokines and chemokines, 11 cell surface receptors and signaling molecules, 7 enzymes and proteases, 11 growth factors and related molecules, and 2 other regulators and proteins (list of proteins seen in Supplementary S2). All 75 protein markers were used to build predictive models as elaborated in Fig.\u0026nbsp;2B.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.4. Comparison of protein levels across groups\u003c/h2\u003e \u003cp\u003eFigure 2B refers to the protein markers with statistically significant fold changes for comparisons between HCL and cases. 30 proteins exhibited significantly distinct levels between HCL and cases (i.e., ARE \u0026amp; CRE \u0026amp; CRT) (full list of proteins with fold change and \u003cem\u003eP\u003c/em\u003e-values in Supplementary S2). Notably, CST5, DNER, SCF, and TWEAK displayed downregulated levels in individuals with schizophrenia, while CCL19, CCL3, CD5, CDCP1, CSF-1, FGF-21, FGF-23, Flt3L, HGF, IL10, IL-10RB, IL-12B, 12-15RA, IL-17C, IL18, IL6, MCP-3, MMP-1, MMP-10, SLAMF1, TGF-alpha, TNF, TNFRSF9, TNFSF14, and VEGFA exhibited upregulated levels. Most of these dysregulated immune biomarkers share functions related to immune regulation, inflammation, and cell signaling. They act on common pathways and networks such as immune cell recruitment and activation, T cell differentiation and proliferation, modulation of cytokine signaling, and cell growth and survival.\u003c/p\u003e \u003cp\u003eFigure 2C refers to the protein markers with statistically significant fold change for comparison between antipsychotic responsive and TRS groups. Five proteins (CCL25, CD5, CST5, MMP-10, and TNFRSF9) showed significant differences between antipsychotic responsive and TRS groups (ARE vs CRT\u0026thinsp;+\u0026thinsp;CRE); CCL25 emerged as the exclusive marker unique to antipsychotic responsiveness.\u003c/p\u003e \u003cp\u003eNo significantly different protein markers were identified between clozapine responsive.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eSubgrouping individuals according to their response to antipsychotic treatment has potential for identifying clinically relevant prognostic markers. This could lead to effective personalised treatments which will improve clinical outcomes in individuals with schizophrenia. However, this remains a significant challenge due to the intricate and multifactorial nature of the condition. To the best of our knowledge, this study marks an early effort to develop a predictive model for categorizing antipsychotic responsiveness.\u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Model prediction\u003c/h2\u003e \u003cp\u003eThere have been numerous attempts to develop predictive models for precision psychiatry using ML and deep learning algorithms. However, most are trained on neuroimaging, speech patterns and digital features \u003csup\u003e\u003cspan additionalcitationids=\"CR22\" citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Only a handful were based on molecular biomarker data \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. The present study was able to achieve a reasonable performance of 0.74 to distinguish people with and without schizophrenia, and 0.88 and 0.78 respectively for differentiating responsiveness to antipsychotics and clozapine. However, the drop in performance for model 3 (differentiating clozapine responsiveness) could be due to small sample size.\u003c/p\u003e \u003cp\u003eRegardless of accuracy and performance, all models still require external validation or practical implementation in clinical settings. So far, very few prediction tools developed in mental health research have progressed into clinical use, with the prediction of medication response in schizophrenia still in its early stages \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Two studies focused on predicting response to medications commonly used in bipolar disorder, namely lithium and quetiapine, utilizing baseline sociodemographic, clinical, and family history information \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Although both studies developed models that performed above chance levels, neither model underwent validation in independent samples. In future research, a crucial strategy will be the integration of biological data for comprehensive analyses. Biological data offers objective measurements of physiological processes or biomarkers, enhancing reliability and reducing variability. Furthermore, the ability to longitudinally monitor changes over time through biological data offers valuable insights into disease progression, treatment efficacy, and individual health trajectories. Therefore, integrating biological data holds significant potential to enhance overall model performance and deepen our understanding of the genetic foundations of mental illnesses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Inflammatory markers\u003c/h2\u003e \u003cp\u003eUsing ML, we identified 18 proteins to be the most impactful features to differentiate healthy individuals from people with schizophrenia. These investigated proteins were involved in inflammatory processes, such as Jak-STAT, NF-κB, mitogen-activated protein kinases (MAPK), RAS, and TNF signaling pathway, cytokine-cytokine, and receptor interaction pathways \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. 11 out of these 18 proteins (CD5, CST5, DNER, IL-10, IL-10RB, IL-12B, IL-15RA, IL-17C, MCP-3, MMP-10, and SCF) exhibited significant difference in protein expression levels between healthy individuals and schizophrenia group.\u003c/p\u003e \u003cp\u003eThere were reports on differential levels of immune biomarkers during different stages of psychosis (prodromal, first-episode psychosis and chronic schizophrenia against healthy individuals). Results from the present study are concordant with reported findings \u0026ndash; higher protein expression levels of commonly studied markers such as CCL23, IL-6, IL-8, IL-10, IL-12, IL-15, TNF and TGF are observed in individuals with schizophrenia when compared to healthy individuals \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. As chemokines, growth factors, cell surface receptors traditionally, received less attention than cytokines, reports of such group of inflammatory markers in psychosis or schizophrenia are limited and often inconsistent. Nonetheless, we observed a general upward trend for similar markers for individuals with schizophrenia in the present study \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSeveral meta-analyses on associations between immunological markers and antipsychotics response were limited to pre- and post-treatment in first episode psychosis \u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. Among numerous immune biomarkers studied, IL-1β, IL-6, and TGF-β were consistently reported to have significant reductions post-antipsychotic treatment. Thus far, only one meta-analysis conducted to evaluate the immune alterations induced by clozapine which reported a down-regulation of IL-6 levels in patients who received clozapine, compared to other antipsychotics \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Although the present study had similar down-regulation of IL-6 (log2 fold change of -0.112) in TRS compared to ARE, this fold change did not meet statistical significance. Out of the 8 immune markers identified via ML, only IL-18 and TNF from this study was associated with response to amisulpride, risperidone, olanzapine, and/or quetiapine \u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. The remaining six markers, CCL25, CST5, MCP-2, MCP-4, PD-L1, and TNFRSF9 require further investigation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.3. The value of ML for classifying pharmacological subtypes\u003c/h2\u003e \u003cp\u003eML methods such as the SVM used in this study can discern subtle signals that may not be identified through traditional statistical testing. Unlike conventional statistical methods like the t-tests, ML methods can model multiple features simultaneously and capture relationships between features, which might be overlooked.\u003c/p\u003e \u003cp\u003eThe SVM was selected for its simplicity, robustness, and its reliance on a good feature set. When coupled with the explainable AI approach SHAP, we were able to improve model interpretability. Based on Fig.\u0026nbsp;3, several statistically significant features had less predictive power compared to features that were not statistically significant based on the mean absolute SHAP value. Hence, this approach can be powerful for biomarker screening. Contemporary methods like ensemble methods (such as bagging and boosting) are recognised for their superior prediction performance score and popularity. However, in this context, they may not be appropriate due to our limited sample and feature size.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Strengths and limitations\u003c/h2\u003e \u003cp\u003eStrengths. The sample population is from a single site in a tertiary healthcare institute, which enabled enrollment of a well-characterised clinical sample and reduce heterogeneity in data and sample collection. Adopting plasma proteins enables high throughout analysis, allowing for the simultaneous identification and measurement of a vast array of proteins.\u003c/p\u003e \u003cp\u003eLimitations. The inflammation panel in the current study is a general list and not tailored for psychosis or mental disorders. We were unable to enroll a drug-na\u0026iuml;ve group to investigate the impact of medication exposure on peripheral immune profiles. Our study lacked longitudinal evaluation, essential for understanding the consistency of immune profiles among pharmacological subtypes, as well as the influence of factors such as exposure to infectious agents or alterations in medications or treatment regimes.\u003c/p\u003e \u003cp\u003eWe demonstrated the efficacy of using ML to predict pharmacological subtypes of schizophrenia based on blood-based inflammatory biomarkers. Our study shows that extracted biomarkers from ML models exhibit different expression levels but do not necessarily attain statistical significance possibly due to limited sample sizes and clinical heterogeneity. Despite these limitations, the functional characterization of ML selected biomarkers is supported by contemporary literature, which validates their associations with biological pathways or processes that are relevant to schizophrenia. Hence, ML-based models can be useful for predicting treatment outcomes and yield meaningful markers for understanding underlying mechanisms.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research is supported by the Singapore Ministry of Health\u0026rsquo;s National Medical Research Council under its MOH-CSAINV17nov-0004.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCredit authorship contribution statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJL designed the study and obtained the funding.\u0026nbsp;JYY and YMS managed the study, supported recruitment and data collection. JYY and SXP analysed the data. AKA assisted with Olink data generation. JYY, SXP, WWBG and JL interpreted the results. JYY and SXP drafted the manuscript. WWBG and JL revised the manuscript. All authors have read and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclarations of Competing Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJL has received honoraria from Sumitomo Pharmaceuticals, Lundbeck Singapore, Otsuka Pharmaceutical and Janssen Pharmaceutical. All other authors declare that they have no conflict of interest.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to express their sincere gratitude to all the participants who participated in this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003ePatel, K. R., Cherian, J., Gohil, K. \u0026amp; Atkinson, D. Schizophrenia: overview and treatment options. \u003cem\u003eP t\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 638-645 (2014).\u003c/li\u003e\n\u003cli\u003eFarooq, S., Agid, O., Foussias, G. \u0026amp; Remington, G. Using treatment response to subtype schizophrenia: proposal for a new paradigm in classification. \u003cem\u003eSchizophr Bull\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 1169-1172, doi:10.1093/schbul/sbt137 (2013).\u003c/li\u003e\n\u003cli\u003eZhu, L.\u003cem\u003e et al.\u003c/em\u003e The machine learning algorithm for the diagnosis of schizophrenia on the basis of gene expression in peripheral blood. \u003cem\u003eNeuroscience letters\u003c/em\u003e \u003cstrong\u003e745\u003c/strong\u003e, 135596, doi:10.1016/j.neulet.2020.135596 (2021).\u003c/li\u003e\n\u003cli\u003eNucifora, F. C., Jr., Woznica, E., Lee, B. J., Cascella, N. \u0026amp; Sawa, A. Treatment resistant schizophrenia: Clinical, biological, and therapeutic perspectives. \u003cem\u003eNeurobiology of disease\u003c/em\u003e \u003cstrong\u003e131\u003c/strong\u003e, 104257, doi:10.1016/j.nbd.2018.08.016 (2019).\u003c/li\u003e\n\u003cli\u003eSchwarz, M. J., M\u0026uuml;ller, N., Riedel, M. \u0026amp; Ackenheil, M. The Th2-hypothesis of schizophrenia: a strategy to identify a subgroup of schizophrenia caused by immune mechanisms. \u003cem\u003eMed Hypotheses\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 483-486, doi:10.1054/mehy.2000.1203 (2001).\u003c/li\u003e\n\u003cli\u003eMiller, B. J. \u0026amp; Goldsmith, D. R. Evaluating the Hypothesis That Schizophrenia Is an Inflammatory Disorder. \u003cem\u003eFocus (American Psychiatric Publishing)\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 391-401, doi:10.1176/appi.focus.20200015 (2020).\u003c/li\u003e\n\u003cli\u003eCortes-Briones, J. A., Tapia-Rivas, N. I., D\u0026apos;Souza, D. C. \u0026amp; Estevez, P. A. Going deep into schizophrenia with artificial intelligence. \u003cem\u003eSchizophrenia research\u003c/em\u003e \u003cstrong\u003e245\u003c/strong\u003e, 122-140, doi:10.1016/j.schres.2021.05.018 (2022).\u003c/li\u003e\n\u003cli\u003eKozyrev, E. A.\u003cem\u003e et al.\u003c/em\u003e Building Predictive Models for Schizophrenia Diagnosis with Peripheral Inflammatory Biomarkers. \u003cem\u003eBiomedicines\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, doi:10.3390/biomedicines11071990 (2023).\u003c/li\u003e\n\u003cli\u003eChilla, G. S., Yeow, L. Y., Chew, Q. H., Sim, K. \u0026amp; Prakash, K. N. B. Machine learning classification of schizophrenia patients and healthy controls using diverse neuroanatomical markers and Ensemble methods. \u003cem\u003eScientific reports\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 2755, doi:10.1038/s41598-022-06651-4 (2022).\u003c/li\u003e\n\u003cli\u003eShim, M., Hwang, H. J., Kim, D. W., Lee, S. H. \u0026amp; Im, C. H. Machine-learning-based diagnosis of schizophrenia using combined sensor-level and source-level EEG features. \u003cem\u003eSchizophrenia research\u003c/em\u003e \u003cstrong\u003e176\u003c/strong\u003e, 314-319, doi:10.1016/j.schres.2016.05.007 (2016).\u003c/li\u003e\n\u003cli\u003eFerrara, M.\u003cem\u003e et al.\u003c/em\u003e Machine Learning and Non-Affective Psychosis: Identification, Differential Diagnosis, and Treatment. \u003cem\u003eCurrent psychiatry reports\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 925-936, doi:10.1007/s11920-022-01399-0 (2022).\u003c/li\u003e\n\u003cli\u003eHaro, J. M.\u003cem\u003e et al.\u003c/em\u003e The Clinical Global Impression-Schizophrenia scale: a simple instrument to measure the diversity of symptoms present in schizophrenia. \u003cem\u003eActa psychiatrica Scandinavica. Supplementum\u003c/em\u003e, 16-23, doi:10.1034/j.1600-0447.107.s416.5.x (2003).\u003c/li\u003e\n\u003cli\u003eLee, J.\u003cem\u003e et al.\u003c/em\u003e Subtyping Schizophrenia by Treatment Response: Antipsychotic Development and the Central Role of Positive Symptoms. \u003cem\u003eCanadian journal of psychiatry. Revue canadienne de psychiatrie\u003c/em\u003e \u003cstrong\u003e60\u003c/strong\u003e, 515-522, doi:10.1177/070674371506001107 (2015).\u003c/li\u003e\n\u003cli\u003eLambert, M., Karow, A., Leucht, S., Schimmelmann, B. G. \u0026amp; Naber, D. Remission in schizophrenia: validity, frequency, predictors, and patients\u0026apos; perspective 5 years later. \u003cem\u003eDialogues in clinical neuroscience\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 393-407, doi:10.31887/DCNS.2010.12.3/mlambert (2010).\u003c/li\u003e\n\u003cli\u003eSiemers, F. M. \u0026amp; Bajorath, J. Differences in learning characteristics between support vector machine and random forest models for compound classification revealed by Shapley value analysis. \u003cem\u003eScientific reports\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 5983, doi:10.1038/s41598-023-33215-x (2023).\u003c/li\u003e\n\u003cli\u003eScott Lundberg, S.-I. L. A Unified Approach to Interpreting Model Predictions. \u003cem\u003earxiv\u003c/em\u003e (2017).\u003c/li\u003e\n\u003cli\u003eGuyon, I., Weston, J., Barnhill, S. \u0026amp; Vapnik, V. Gene Selection for Cancer Classification using Support Vector Machines. \u003cem\u003eMachine Learning\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 389-422, doi:10.1023/A:1012487302797 (2002).\u003c/li\u003e\n\u003cli\u003eLi, Z., Xie, W. \u0026amp; Liu, T. Efficient feature selection and classification for microarray data. \u003cem\u003ePloS one\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, e0202167, doi:10.1371/journal.pone.0202167 (2018).\u003c/li\u003e\n\u003cli\u003eWang, Z., Meng, Z. \u0026amp; Chen, C. Screening of potential biomarkers in peripheral blood of patients with depression based on weighted gene co-expression network analysis and machine learning algorithms. \u003cem\u003eFrontiers in psychiatry\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1009911, doi:10.3389/fpsyt.2022.1009911 (2022).\u003c/li\u003e\n\u003cli\u003eLundberg, S. M. \u0026amp; Lee, S.-I. A unified approach to interpreting model predictions. \u003cem\u003eAdvances in neural information processing systems\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e (2017).\u003c/li\u003e\n\u003cli\u003eRezaii, N., Walker, E. \u0026amp; Wolff, P. A machine learning approach to predicting psychosis using semantic density and latent content analysis. \u003cem\u003enpj Schizophrenia\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 9, doi:10.1038/s41537-019-0077-9 (2019).\u003c/li\u003e\n\u003cli\u003eNenning, K.-H. \u0026amp; Langs, G. Machine learning in neuroimaging: from research to clinical practice. \u003cem\u003eDie Radiologie\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 1-10, doi:10.1007/s00117-022-01051-1 (2022).\u003c/li\u003e\n\u003cli\u003eBenoit, J., Onyeaka, H., Keshavan, M. \u0026amp; Torous, J. Systematic Review of Digital Phenotyping and Machine Learning in Psychosis Spectrum Illnesses. \u003cem\u003eHarvard Review of Psychiatry\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 296-304, doi:10.1097/hrp.0000000000000268 (2020).\u003c/li\u003e\n\u003cli\u003eChekroud, A. M.\u003cem\u003e et al.\u003c/em\u003e The promise of machine learning in predicting treatment outcomes in psychiatry. \u003cem\u003eWorld psychiatry : official journal of the World Psychiatric Association (WPA)\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 154-170, doi:10.1002/wps.20882 (2021).\u003c/li\u003e\n\u003cli\u003eNunes, A.\u003cem\u003e et al.\u003c/em\u003e Prediction of lithium response using clinical data. \u003cem\u003eActa psychiatrica Scandinavica\u003c/em\u003e \u003cstrong\u003e141\u003c/strong\u003e, 131-141, doi:10.1111/acps.13122 (2020).\u003c/li\u003e\n\u003cli\u003eKim, T. T.\u003cem\u003e et al.\u003c/em\u003e Predictive modeling for response to lithium and quetiapine in bipolar disorder. \u003cem\u003eBipolar disorders\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 428-436, doi:10.1111/bdi.12752 (2019).\u003c/li\u003e\n\u003cli\u003eSchnorr, I.\u003cem\u003e et al.\u003c/em\u003e Inflammatory biotype of ADHD is linked to chronic stress: a data-driven analysis of the inflammatory proteome. \u003cem\u003eTranslational psychiatry\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 37, doi:10.1038/s41398-023-02729-3 (2024).\u003c/li\u003e\n\u003cli\u003eMomtazmanesh, S., Zare-Shahabadi, A. \u0026amp; Rezaei, N. Cytokine Alterations in Schizophrenia: An Updated Review. \u003cem\u003eFrontiers in psychiatry\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 892, doi:10.3389/fpsyt.2019.00892 (2019).\u003c/li\u003e\n\u003cli\u003eKorhonen, L.\u003cem\u003e et al.\u003c/em\u003e Multivariate analyses of immune markers reveal increases in plasma EN-RAGE in first-episode psychosis patients. \u003cem\u003eTranslational psychiatry\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 326, doi:10.1038/s41398-023-02627-8 (2023).\u003c/li\u003e\n\u003cli\u003eErmakov, E. A., Mednova, I. A., Boiko, A. S., Buneva, V. N. \u0026amp; Ivanova, S. A. Chemokine Dysregulation and Neuroinflammation in Schizophrenia: A Systematic Review. \u003cem\u003eInternational journal of molecular sciences\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, doi:10.3390/ijms24032215 (2023).\u003c/li\u003e\n\u003cli\u003eLi, X., Hu, S. \u0026amp; Liu, P. Vascular-related biomarkers in psychosis: a systematic review and meta-analysis. \u003cem\u003eFrontiers in psychiatry\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1241422, doi:10.3389/fpsyt.2023.1241422 (2023).\u003c/li\u003e\n\u003cli\u003eMarcinowicz, P.\u003cem\u003e et al.\u003c/em\u003e A Meta-Analysis of the Influence of Antipsychotics on Cytokines Levels in First Episode Psychosis. \u003cem\u003eJournal of clinical medicine\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, doi:10.3390/jcm10112488 (2021).\u003c/li\u003e\n\u003cli\u003eOrbe, E. B. \u0026amp; Benros, M. E. Immunological Biomarkers as Predictors of Treatment Response in Psychotic Disorders. \u003cem\u003eJournal of personalized medicine\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, doi:10.3390/jpm13091382 (2023).\u003c/li\u003e\n\u003cli\u003eMartins, P. L. B.\u003cem\u003e et al.\u003c/em\u003e Immunoinflammatory and oxidative alterations in subjects with schizophrenia under clozapine: A meta-analysis. \u003cem\u003eEuropean neuropsychopharmacology : the journal of the European College of Neuropsychopharmacology\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e, 82-95, doi:10.1016/j.euroneuro.2023.04.003 (2023).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"translational-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"tp","sideBox":"Learn more about [Translational Psychiatry](http://www.nature.com/tp/)","snPcode":"41398","submissionUrl":"https://mts-tp.nature.com/cgi-bin/main.plex","title":"Translational Psychiatry","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Artificial Intelligence (AI), Inflammation, Machine Learning (ML), Peripheral Biomarkers, Treatment Resistant Schizophrenia (TRS)","lastPublishedDoi":"10.21203/rs.3.rs-4604742/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4604742/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eIn this research study, we apply machine learning techniques to navigate the multifaceted landscape of schizophrenia. Our method entails the development of predictive models, emphasizing peripheral inflammatory biomarkers, which are classified into treatment response subgroups: antipsychotic-responsive, clozapine-responsive, and clozapine-resistant.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThe study comprises 146 schizophrenia patients (49 antipsychotics-responsive, 68 clozapine-responsive, 29 clozapine-resistant) and 49 healthy controls. Protein levels of immune biomarkers were quantified using the Olink Target 96 Inflammation Panel. To predict labels, a support vector machine classifier is trained on the Olink data matrix and evaluated via leave-one-out cross-validation. Associated protein biomarkers are identified via recursive feature elimination.\u003c/p\u003e\u003ch2\u003eFindings\u003c/h2\u003e \u003cp\u003eWe constructed three separate predictive models for binary classification: one to discern healthy controls from individuals with schizophrenia (AUC\u0026thinsp;=\u0026thinsp;0.74), another to differentiate individuals who were responsive to antipsychotics (AUC\u0026thinsp;=\u0026thinsp;0.88), and a third to distinguish treatment-resistant individuals (AUC\u0026thinsp;=\u0026thinsp;0.78). Employing machine learning techniques, we identified features capable of distinguishing between treatment response subgroups.\u003c/p\u003e\u003ch2\u003eInterpretation\u003c/h2\u003e \u003cp\u003eIn this study, support vector machine demonstrates the power of machine learning to uncover subtle signals often overlooked by traditional statistics. Unlike t-tests, it handles multiple features simultaneously, capturing complex data relationships. Chosen for simplicity, robustness, and reliance on strong feature sets, its integration with artificial intelligence techniques like SHapely Additive exPlanations enhances model interpretability, especially for biomarker screening. This study highlights the potential of integrating machine learning techniques in clinical practice. Not only does it deepen our understanding of schizophrenia's heterogeneity, but it also holds promise for enhancing predictive accuracy, thereby facilitating more targeted and effective interventions in the treatment of this complex mental health disorder.\u003c/p\u003e","manuscriptTitle":"Predicting antipsychotic responsiveness using a machine learning classifier trained on plasma levels of inflammatory markers in schizophrenia","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-10 11:33:23","doi":"10.21203/rs.3.rs-4604742/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"revise","date":"2024-08-19T14:35:44+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-07-30T08:25:30+00:00","index":3,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-07-22T13:21:05+00:00","index":2,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-07-22T04:41:32+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-07-15T13:33:02+00:00","index":3,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-07-10T09:25:10+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-07-09T23:16:46+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2024-07-09T21:45:00+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-06-24T12:05:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"Translational Psychiatry","date":"2024-06-24T09:29:00+00:00","index":"","fulltext":""},{"type":"checksFailed","content":"","date":"2024-06-20T10:09:51+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-06-19T09:15:30+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"translational-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"tp","sideBox":"Learn more about [Translational Psychiatry](http://www.nature.com/tp/)","snPcode":"41398","submissionUrl":"https://mts-tp.nature.com/cgi-bin/main.plex","title":"Translational Psychiatry","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"36d9d356-1787-4405-a0f1-07691fa2d34d","owner":[],"postedDate":"August 10th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":34367133,"name":"Health sciences/Biomarkers/Predictive markers"},{"id":34367134,"name":"Health sciences/Diseases/Psychiatric disorders/Schizophrenia"}],"tags":[],"updatedAt":"2025-02-15T08:05:40+00:00","versionOfRecord":{"articleIdentity":"rs-4604742","link":"https://doi.org/10.1038/s41398-025-03264-z","journal":{"identity":"translational-psychiatry","isVorOnly":false,"title":"Translational Psychiatry"},"publishedOn":"2025-02-14 05:00:00","publishedOnDateReadable":"February 14th, 2025"},"versionCreatedAt":"2024-08-10 11:33:23","video":"","vorDoi":"10.1038/s41398-025-03264-z","vorDoiUrl":"https://doi.org/10.1038/s41398-025-03264-z","workflowStages":[]},"version":"v1","identity":"rs-4604742","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4604742","identity":"rs-4604742","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}